EPFL Logo CENTER FOR
DIGITAL TRUST

Center for Digital Trust's Showcase

Welcome to the Center for Digital Trust (C4DT)'s showcase, a hub of cutting‑edge research and innovation in digital trust from EPFL’s leading laboratories. Here you’ll find the most recent projects that advance security, privacy, and trustworthy computing, selected by the C4DT Factory team.

For more details about the Center for Digital Trust, visit our main site. The Factory team shares its latest news and insights on our blog. If you need to get in touch with the Factory team, you can reach us by emailing factory@c4dt.org.

View all Projects

Filter projects
Reset Filters

Anyway

Self-hosted AI platform for private, on-premise LLM inference

Anyway orchestrates on-premise hardware into a distributed inference cluster supporting any Hugging Face model. It deploys as a standalone app, Docker image, or Kubernetes pod and provides an OpenAI-compatible REST API endpoint, giving organisations full model control and fixed operational costs with no per-token fees.

Rachid Guerraoui

Distributed Computing Lab

Maturity
Support
C4DT
Active
Lab
Unknown
2026 ProposalDecentralizedLarge Language ModelMachine Learning
PapersArticles
Project Homepage
Contact

PrivacyShield

BLE beacon relay system to prevent unwanted AirTag tracking using ESP32 relays.

System countering unsolicited BLE tracker-based surveillance by relay-confusing AirTag and similar devices. Architecture: Flask server, modified AirGuard Android app, ESP32 BLE relay firmware, and optional Sniffle-based sniffer tools. Fully open-source under GPLv3/MIT.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Active
2026 ProposalBluetoothInternet of Things
PapersArticles
Source Code

HalluHard

Hard multi-turn hallucination benchmark for evaluating language models across domains.

Multi-turn hallucination benchmark evaluating LLMs across diverse domains. Installed via Pixi; scripts provided for response generation, claim-based web-scraping judgment (or coding_direct mode), and report creation. Supports multiple models and CLI configuration. Designed to maximize hallucination elicitation difficulty.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
2026 ProposalAI SafetyBenchmarkLarge Language Model
PapersArticles
Source Code

Votegral

End-to-end verifiable, coercion-resistant voting system using TRIP for in-person registration.

End-to-end verifiable, coercion-resistant voting system using TRIP for in-person registration. Implements E2E verification and coercion resistance via randomized ballot shuffling and re-encryption. Includes simulation framework, Docker/local build, configurable hardware and shuffle modes, and performance metrics generation.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
2026 ProposalProtocol
PapersArticles
Source Code

DebateGPT

Code and data for studying the conversational persuasiveness of GPT-4.

Code and data for studying conversational persuasiveness of GPT-4 vs. human debaters, published in Nature Human Behaviour. Includes scripts for data processing, linguistic feature extraction, and statistical analysis. Reproduces all experiments from the associated arXiv paper.

Robert West

Data Science Lab

Maturity
Support
C4DT
Inactive
Lab
Active
2026 ProposalLarge Language Model
PapersArticles
Source Code

TiME

Training tiny monolingual language models via knowledge distillation from larger models.

Knowledge distillation pipeline for training tiny monolingual encoders from XLM-R or HPLT teacher models. Supports downstream evaluation for POS tagging, lemmatization, dependency parsing, and NER. Includes checkpoint selection scripts, distillation pipelines, and evaluation tooling.

Robert West

Data Science Lab

Maturity
Support
C4DT
Inactive
Lab
Active
2026 ProposalDeep Neural NetworksNatural Language
PapersArticles
Source Code

Truman

Constructs device behavior models from OS drivers for virtual-device fuzzing.

Framework for constructing device behavior models from OS device drivers, capturing inter- and intra-message dependencies. Generates dependency-aware seeds for virtual-device fuzzing via dynamic binary instrumentation (DBI). Docker-based build, fuzzing, and coverage collection pipelines provided.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Active
2026 ProposalFuzzing
PapersArticles
Source Code

SPD

Single-pass detection classifier for jailbreaking inputs in large language models.

Single-pass jailbreak detection classifier for LLMs extending JailbreakBench. Trained on 4,000+ samples from GCG, AutoDAN, and PAIR attack methods. TMLR 2025 paper. Provides training and evaluation scripts and a unified detection interface.

Volkan Cevher

Laboratory for Information and Inference Systems

Maturity
Support
C4DT
Inactive
Lab
Unknown
2026 ProposalAI SafetyAttackLarge Language Model
PapersArticles
Source Code

Reflecta

Reflection-based, language-agnostic fuzzer for scripting languages.

Language-agnostic fuzzer for scripting language interpreters leveraging reflection APIs to generate valid programs. Supports Ruby, MRuby, CPython, MicroPython, PHP, and V8. Dockerized with scripts for campaign launch, coverage collection, and result visualization. Compatible with reflecta, nautilus, and polyglot fuzzers.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
2026 ProposalFuzzing
PapersArticles
Source Code

arkworks

Rust ecosystem for zkSNARK programming

Arkworks provides efficient implementations of components required for zkSNARK applications. It includes libraries for finite fields, elliptic curves, R1CS constraints, and SNARK proving systems like Groth16 and Marlin. It defines interfaces for relations like R1CS, and for building SNARK circuits using gadgets and subcircuits.

Alessandro Chiesa

Laboratory for Computation Security

Maturity
PrototypeIntermediateMature
Support
C4DT
Active
Lab
Active
Zero-Knowledge Proofs
PapersArticles
Project Homepage
Source Code

d-voting

A decentralized, fully anonymous voting system for EPFL-internal elections.

D-Voting is built on the Dela distributed ledger and uses a threshold ElGamal encryption scheme for ballot confidentiality combined with Neff shuffles for anonymization. Decryption is performed collectively by a set of servers using distributed key generation (DKG), ensuring that no single server can decrypt or de-anonymize ballots on its own.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Active
Lab
Active
AnonymityByzantine ResilienceDecentralized
PapersArticles
Project Homepage
Source Code
Contact

Secure and Privacy-Preserving Credentials for E-ID

Research techniques to improve privacy when using electronic identities

Today's identity security faces challenges like misuse and tracking. Our goal is to enable secure, anonymous, unlinkable E-ID interactions by researching novel cryptographic algorithms. This boosts user trust, creates new business opportunities, and cuts financial losses after data breaches.

Jean-Pierre Hubaux

Center For Digital Trust

Maturity
PrototypeIntermediateMature
Support
C4DT
Active
Lab
Active
Access ControlDecentralizedZero-Knowledge Proofs
PapersArticles
Project Homepage
Source Code

Showcase v2.0

Version 2.0 of the C4DT's showcase.

A statically generated TypeScript web application that aggregates and renders YAML-described research project metadata from multiple EPFL labs. It provides filtering, search, and structured display of project maturity, code repositories, and related publications.

Jean-Pierre Hubaux

Center For Digital Trust

Maturity
PrototypeIntermediateMature
Support
C4DT
Active
Lab
Active
Development
PapersArticles
Source Code

Disco

Decentralized Collaborative Machine Learning

A collection of objects and routines to help you develop your own distributed machine learning algorithm. It has two modes, federated, where a server is keeping track of the model, and decentralized, where each participant has their own local model. You can train your model directly in the browser, allowing it to run on a variety of hardware. Many extensions are in the pipeline, such as adding Byzantine resistance or training without revealing the model. Compared to other solutions, Disco is bridging the gap between people creating machine learning models and people having relevant data. By running without installation nor complex configuration, anyone can help. It's also a breeding ground for new machine learning technologies.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Active
Lab
Active
DecentralizedDistributed LearningTensorFlow
PapersArticles
Project Homepage
Source Code

Orchestrated Artifact Evaluation

Automatically evaluating code artifacts from research papers

While there has been a strong push for adding reliable code artifacts to research papers since 2018, the process of evaluating them is still manual and often very tedious. Orchard proposes a standardized framework based on Docker to automatically evaluate those artifacts.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Active
Lab
Unknown
BenchmarkCode Analysis
PapersArticles

Combattre le phishing: quelles innovations apporter ?

Combating phishing through vulnerability analysis, detection tools, and best practice guidelines.

This project addresses phishing threats through the analysis of existing countermeasures, identification of vulnerabilities, and development of innovative solutions including best practice guidelines and tools for email configuration testing and malicious link detection. The initiative enhances organizational and societal resilience against phishing attacks through knowledge dissemination and fosters collaboration among stakeholders.

Jean-Pierre Hubaux

Center For Digital Trust

Maturity
PrototypeIntermediateMature
Support
C4DT
Active
Lab
Unknown
AttackCloudExploitProtection
PapersArticles
Project Homepage

Sécurisation de Digiinov Hmi en adéquation avec le Cyber Resilience Act (CRA)

Securing industrial HMI software for EU Cyber-Resilience-Act compliance

This project integrates state-of-the-art software security measures into Digiinov HMI, a platform bridging Operational Technology (OT) and Information Technology (IT) for industrial Human-Machine Interfaces. The collaboration with HEIG-VD and EPFL focuses on achieving compliance with the EU Cyber-Resilience-Act (CRA) requirements before its 2027 enforcement. The security enhancements target the software development lifecycle and maintenance processes for embedded industrial control systems.

Jean-Pierre Hubaux

Center For Digital Trust

Maturity
Support
C4DT
Active
Lab
Unknown
Code AnalysisLarge Language Model
PapersArticles

Matrix@EPFL

The Matrix server run for the EPFL community by the EPFL Center for Imaging with help from the EPFL Center for Digital Trust.

Matrix@EPFL is a self-hosted deployment of the Matrix open federation protocol, running a Synapse homeserver for the EPFL community. It provides end-to-end encrypted messaging, room-based collaboration, and federation with the broader Matrix network, serving as a privacy-respecting alternative to proprietary chat platforms.

Jean-Pierre Hubaux

Center For Digital Trust

Maturity
Support
C4DT
Active
Lab
Unknown
DecentralizedEncryptionProtection
PapersArticles
Project Homepage

Lattigo

A lattice-based multiparty homomorphic encryption library in Go

Lattigo is a Go module for Multiparty Homomorphic Encryption, that implements Ring-Learning-With-Errors-based homomorphic encryption primitives. It comprises highly-efficient RNS versions of the BFV and CKKS schemes and their multiparty (N-out-of-N-threshold) counterparts. It also implements a dense-key and sparse-key Bootstrapping procedure for RNS-CKKS. The Lattigo library was originally exclusively developed by the EPFL Laboratory for Data Security until its version 2.4.0. Starting with the release of version 3.0.0, Lattigo is maintained and supported by Tune Insight SA. Lattigo 2.0.0 has been code-reviewed by ELCA in November 2020 and, within the allocated time for the code review, no critical or high-risk issues were found.

Jean-Pierre Hubaux

Laboratory for Data Security

Maturity
PrototypeIntermediateMature
Support
C4DT
Retired
Lab
Active
Homomorphic EncryptionSecure Multi-Party Computation
PapersArticles
Source Code

MetaWearS

Meta-learning framework for few-shot adaptation in wearable health monitoring.

Meta-learning framework for few-shot adaptation in wearable health monitoring. Pre-trains on large datasets with prototypical network structure; fine-tunes on few shots per user. Lightweight update mechanism conserves battery. Provides code, Jupyter notebooks, and hardware implementation details.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
Deep Neural NetworksInternet of ThingsMedicalTime SeriesWearable
PapersArticles
Source Code

nwsim

Network simulation suite integrating ns-2 with pFabric, SIRD, DCTCP, and Homa.

ns-2 simulation suite extended with pFabric, SIRD, DCTCP, and Homa transport protocols, plus OMNET integration. Provides configuration scripts for incast and random traffic experiments and output processing tools for queue occupancy and application metrics.

Edouard Bugnion

Data Center Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
NetworkSimulator
PapersArticles
Source Code

Miralis

Virtual firmware monitor for RISC-V with formal verification via Kani.

Virtual firmware monitor for RISC-V running below M-mode firmware, providing isolation and policy enforcement via sandbox and Keystone policies. Includes a Rust RISC-V ISA translator, benchmark scripts, and formal verification via Kani. Supports Docker and manual setup. SOSP'25 artifact.

Edouard Bugnion

Data Center Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
Formal MethodsLow-LevelRISC-V
PapersArticles
Source Code

PEPO

Preference alignment framework for language models using DPO and RLHF techniques.

Preference alignment framework for LLMs implementing DPO, RLHF, and related techniques. Uses Hydra for configuration management; supports SLURM cluster scheduling, CUDA systems, training and evaluation pipelines, and environment variable management. Pre-commit hooks included for code quality.

Volkan Cevher

Laboratory for Information and Inference Systems

Maturity
Support
C4DT
Inactive
Lab
Active
AI SafetyLarge Language ModelOptimization
PapersArticles
Source Code

TiMoE

Modular framework for temporally grounded language models with time-aware expert routing.

Modular Mixture-of-Experts (MoE) framework for temporally grounded LLMs. Trains separate GPT experts on two-year data slices; routes queries to experts up to the query timestamp. Aggregates outputs via equal weighting, learned routing, or joint co-adaptation. Includes repo structure and quickstart instructions.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
Large Language Model
PapersArticles
Source Code

NEX

Full-stack simulation framework running unmodified software while modelling hardware accelerators.

Full-stack simulation framework running unmodified software stacks on real CPUs while modeling hardware accelerators (VTA, JPEG, Protoacc) via RTL simulation or DSim. Supports multi-accelerator setups, DMA and interconnect latency modeling, and epoch-based scheduling. Integrates with the SimBricks modular simulation platform.

George Candea

Dependable Systems Lab

Maturity
Support
C4DT
Inactive
Lab
Active
Simulator
PapersArticles
Source Code

Timeseries Saliency Maps

Cross-domain saliency maps library for time-series model explanation in PyTorch and TensorFlow.

Library implementing cross-domain saliency maps for time-series model explainability in PyTorch and TensorFlow. Supports time, frequency, and ICA domains via invertible differentiable transforms. Requires no model retraining; extensible to custom transforms. Pip-installable with ready-to-run notebooks.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
PyTorchTensorFlowTime SeriesWearable
PapersArticles
Source Code

BoostCD

Information extraction system with training, inference, and evaluation pipeline.

End-to-end information extraction system with training, inference, and evaluation pipelines. Uses a conda environment and Hugging Face datasets; supports training on custom data, batched inference, and prediction postprocessing. Published at an NLP conference.

Robert West

Data Science Lab

Maturity
Support
C4DT
Inactive
Lab
Active
Natural Language
PapersArticles
Source Code

libErator

Automatically generates fuzzing drivers for libraries by extracting constraints.

Framework for automated fuzzing driver synthesis for C/C++ libraries. Uses static analysis to extract API constraints and pre/post-conditions, then generates conforming drivers for libfuzzer. Includes a customized libfuzzer build and tooling to add new fuzzing targets and run experiments.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Active
FuzzingStatic Analysis
PapersArticles
Source Code

GP TA Emulator

GlobalPlatform Trusted Application emulator with GDB and fuzzing support.

Emulator for GlobalPlatform-compliant Trusted Applications (TAs) enabling development, debugging (GDB support), and fuzzing without physical TEE hardware. Provides Docker setup, PoC build instructions, Android NDK cross-compilation, and fuzzing harness with crash handling and replay.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Active
FuzzingTrusted Execution Environment
PapersArticles
Source Code

NoiseDiff

Dark Noise Diffusion for synthetic low-light noise generation and image denoising.

Dark Noise Diffusion model for synthetic low-light noise generation and paired image denoising. Trained on SID and ELD datasets; uses diffusion-based noise simulation to augment training data. Provides pretrained models, training/testing scripts, shading resources, and optional synthetic data generation pipeline.

Sabine Süsstrunk

Image and Visual Representation Lab

Maturity
Support
C4DT
Inactive
Lab
Active
Deep Neural NetworksImages
PapersArticles
Source Code

zip2zip

High-performance LZW compression library with Python bindings for adaptive vocabularies.

High-performance C LZW compression library with Python bindings (cffi/ctypes). Supports adaptive vocabularies for LLM tokenization pipelines; includes a novel LZW variant that decompresses LLM- generated sequences without requiring a perfectly encoded input stream. Install via pip or from source.

Robert West

Data Science Lab

Maturity
Support
C4DT
Inactive
Lab
Active
Large Language Model
PapersArticles
Project Homepage
Source Code

Tandem Training

Code and data for tandem training of language models.

Code and data for the "Tandem Training for Language Models" paper (arXiv:2510.13551). Implements coordinated joint training of language model pairs. Pre-trained models available on Hugging Face. Authors: West, Anderson, Kamar, Horvitz.

Robert West

Data Science Lab

Maturity
Support
C4DT
Inactive
Lab
Active
Large Language Model
PapersArticles
Source Code

WF4NYM

Website fingerprinting and flow-correlation attacks on the Nym mixnet.

Artifact for website fingerprinting and flow-correlation attacks on the Nym mixnet and Tor. Evaluates the WTF4NYM defense, analyzes feature importance, and measures overhead. Implemented in Python 3.9/3.10 on Linux; includes code, datasets, and full reproduction instructions.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
AnonymityAttackFingerprinting
PapersArticles
Source Code

Scion Optimizer

Norm-constrained optimizer with multiple norm implementations for neural network training.

Norm-constrained optimizer for neural network training with reference implementations of Scion and ScionLight supporting multiple norms (spectral, Frobenius, etc.). Includes hyperparameter guidance, nanoGPT/MLP/CNN example configurations, and notes on a recent momentum initialization change.

Volkan Cevher

Laboratory for Information and Inference Systems

Maturity
Support
C4DT
Inactive
Lab
Active
Deep Neural NetworksOptimization
PapersArticles
Source Code

LEAF

Levenshtein Efficient Adversarial Finetuning for creating a robust CLIP text encoder.

Levenshtein Efficient Adversarial Finetuning (LEAF) for a robust CLIP text encoder resilient to character-level perturbations. Combined with FARE (robust image encoder) for dual-domain robustness. Pre-trained models on HuggingFace; provides training and evaluation procedures.

Volkan Cevher

Laboratory for Information and Inference Systems

Maturity
Support
C4DT
Inactive
Lab
Active
ImagesNatural LanguageRobustness
PapersArticles
Source Code

CHAMELEON

Data-mixing framework using Kernel Ridge Leverage Scores for LLM pretraining and finetuning.

Data-mixing framework using Kernel Ridge Leverage Scores (KRLS) for domain weighting in LLM pretraining and finetuning. Computes domain embeddings and weights to optimize universal generalization and transferability. Integrates with existing training pipelines via simple scripting.

Volkan Cevher

Laboratory for Information and Inference Systems

Maturity
Support
C4DT
Inactive
Lab
Active
Large Language Model
PapersArticles
Source Code

NBL

Neural Block Linearization replacing self-attention with LMMSE approximations for faster LLM inference.

LLM inference acceleration via selective replacement of self-attention with linear LMMSE approximations. CCA-based error bounding identifies low-error layers without fine-tuning. Achieves ~32% speed-up on Llama-8B with <1% accuracy loss. Supports quantization, LoRA fine-tuning, and speculative decoding.

Volkan Cevher

Laboratory for Information and Inference Systems

Maturity
Support
C4DT
Inactive
Lab
Active
Deep Neural NetworksLarge Language Model
PapersArticles
Source Code

ClippedScion

Reference implementation of ClippedScion optimizers supporting multiple norms.

Reference implementation of ClippedScion and UnconstrainedClippedScion optimizers extending Scion with gradient clipping. Supports multiple norms and hyperparameter configurations for nanoGPT, CNN, and DeiT training with example setups and citation information.

Volkan Cevher

Laboratory for Information and Inference Systems

Maturity
Support
C4DT
Inactive
Lab
Active
Deep Neural NetworksOptimization
PapersArticles
Source Code

Gradient-Normalized Smoothness

Algorithms for optimization with approximate Hessians using gradient-normalized smoothness.

Implementation of the gradient-normalized smoothness (GNS) analysis framework for optimization with approximate Hessians. Provides algorithms, oracle functions, Hessian approximation routines, and utilities for convex and non-convex settings. Includes quick-start instructions and examples reproducing paper results.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
Optimization
PapersArticles
Source Code

LLM Optimizer Benchmark

Benchmark for comparing optimizers for LLM pretraining on Llama and MoE models.

Benchmarking suite for comparing optimizers (Adam, SGD, Muon, Scion, etc.) on LLM pretraining using Llama and MoE architectures. Supports configurable training parameters, WandB logging, multi-GPU and CPU setups, and guidelines for extending to new models or datasets.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
BenchmarkLarge Language ModelOptimization
PapersArticles
Source Code

MTBBench

Benchmark for evaluating multimodal LLM reasoning in complex oncology clinical decision-making scenarios

MTBBench is a benchmark for evaluating multimodal LLM reasoning in oncology, covering two core challenges: multimodal integration (pathology, genomics, radiology) and longitudinal reasoning across patient timelines. It includes agentic tasks requiring interaction with external foundation-model-based tools such as TRIDENT for pathology and DrugBank for pharmacology.

Charlotte Bunne

Artificial Intelligence in Molecular Medicine

Maturity
Support
C4DT
Inactive
Lab
Active
BenchmarkLarge Language ModelMachine LearningMedical
PapersArticles
Source Code

NASS

Dynamic binary instrumentation tool for fuzzing Android native system services.

Dynamic binary instrumentation (DBI) framework for discovering and fuzzing Android native system services. Requires a rooted device with ADB and NDK r26. Provides scripts for interface extraction, seed generation, fuzzing campaigns, crash triage, and evaluation against COTS and FANS baselines. Docker setup available.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Active
BinaryFuzzing
PapersArticles
Source Code

PHEE

Posit-enabled X-HEEP RISC-V platform integrating the Coprosit arithmetic coprocessor.

X-HEEP RISC-V SoC extended with the Coprosit posit coprocessor via CV-X-IF, targeting the cv32e40px CPU. Supports configurable posit sizes and quire operations. Provides QuestaSim simulation, PYNQ-Z2 FPGA synthesis, vendor update management, and test commands.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
HardwareLow-LevelRISC-V
PapersArticles
Source Code

Compigra

MLIR-based compilation tool transforming C/C++ code into CGRA assembly.

MLIR-based compilation tool transforming C/C++ to CGRA assembly via custom IR dialects, front-end optimizations, and a back-end ILP mapper requiring Gurobi. Build with CMake; run via compigra-opt; results verifiable with CGRA-simulator or RTL tools.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
CompilerLow-Level
PapersArticles
Source Code

AquaCast

Transformer-based time-series forecasting framework for urban water dynamics.

Research framework for multi-variate, multi-horizon time-series forecasting of urban water dynamics using transformer architectures. Includes synthetic datasets, Docker setup, and experiment scripts. Real operational data unavailable due to NDAs. References an arXiv preprint.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
Deep Neural NetworksPredictorTime Series
PapersArticles
Source Code

Mage

Remote memory system for scalable memory-intensive applications with asynchronous decoupling.

Remote memory system for memory-intensive workloads using asynchronous decoupling, pipelined execution, and contention avoidance at the LibOS layer, implemented on top of RDMA. Artifact includes kernel module, LibOS, baseline comparisons (kernel swap, Fastswap), and full evaluation scripts.

Sanidhya Kashyap

Robust Scalable Systems Software Lab

Maturity
Support
C4DT
Inactive
Lab
Active
ConcurrencyLow-LevelServer
PapersArticles
Source Code

Fold

Rust framework for building dynamic linkers with syscall filtering and function hooks.

Rust framework for constructing custom dynamic linkers with syscall filtering, inter-module communication, and invisible function-hook injection. Includes quick-start guide, required toolchain, VSCode configuration, and example projects demonstrating core capabilities.

Edouard Bugnion

Data Center Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
Low-Level
PapersArticles
Source Code

Lorikeet

Scalafix-based code quality feedback tool with custom rule definitions via query patterns.

Scalafix-based linting and automated grading framework with custom rule definitions via semantic query patterns and rewrite templates. Supports batch student submission checking, formatting feedback, and developer assistance. Scripts included for testing, publishing, and evaluation.

Clément Pit-Claudel

Systems and Formalisms Lab

Maturity
Support
C4DT
Inactive
Lab
Active
Code Analysis
PapersArticles
Source Code

Grounded Arithmetic

Prototype of Grounded Arithmetic logic formalized in Isabelle/Pure.

Isabelle/Pure prototype of the Grounded Arithmetic (GA) logic system. Features a complete axiomatization, arithmetic functions, Cantor tuple constructions, a manually encoded List datatype, and the accompanying B.Sc. thesis PDF and source files.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
Formal Methods
PapersArticles
Source Code

Droidot

Framework to fuzz Android app native libraries on arm64.

Fuzzing framework for Android app native libraries (arm64) using static analysis to extract function signatures and generate fuzzing harnesses. Workflow: Docker setup → APK loading → emulator setup → static analysis → harness generation → fuzzing → crash triage. Includes emulator, fuzzing, harness, and static analysis tooling.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Active
BinaryFuzzing
PapersArticles
Source Code

SYSYPHUZZ

Syzkaller-based kernel fuzzer targeting low-frequency code regions with coverage-guided scheduling.

Kernel fuzzer extending Syzkaller with selective task scheduling to prioritize low-frequency basic blocks and context-preserving mutations. Introduces warm-up mode with basic-block hit-count tracking via kcov instrumentation. Achieves better coverage of rare but bug-prone kernel code paths.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Active
Fuzzing
PapersArticles
Source Code

Beanpod TEE Pwn

Proof-of-concept exploit for the keyinstall Trusted Application on Xiaomi Redmi Note 11s 5G.

PoC exploit targeting the keyinstall Trusted Application in the Beanpod TEE on Xiaomi Redmi Note 11s 5G. Demonstrates mapping of secure monitor memory via a vulnerability in the TA's key installation flow. Requires rooted device; provides kernel module and verification steps.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Active
AttackExploitTrusted Execution Environment
PapersArticles
Source Code

HEEPidermis

X-HEEP-based SoC for GSR and impedance recording with DMA and level-crossing support.

X-HEEP-based SoC for galvanic skin response (GSR) and impedance recording using an iDAC and 14-bit VCO-ADC. Supports DMA, digital Level-Crossing detection, and JTAG/SPI/UART interfaces. Board preparation, programming modes, and RTL simulation steps documented in the README.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
HardwareInternet of ThingsLow-LevelMedicalRISC-VWearable
PapersArticles
Source Code

HEEPidermis Breakout Board

Hardware breakout board for HEEPidermis with connector pinout documentation.

Hardware breakout board for the HEEPidermis SoC providing accessible connector pinouts for integration and testing. Includes a connector pinout spreadsheet and diagram link for hardware engineers.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
HardwareWearable
PapersArticles
Source Code

CHEEP Boards

Modular PCB suite with configurable power supplies and current measurement modules.

Modular PCB suite with a main board (FTDI, power, peripherals) and 150-pin connector for custom breakout boards. Chewing-gum modules add configurable voltage supplies or current measurement. I²C multiplexers manage address conflicts. Open-source with extension APIs for custom module development.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
Hardware
PapersArticles
Source Code

Veritas

Syzkaller-based fuzzer with specification-based oracle for detecting eBPF misbehavior.

Syzkaller-based kernel fuzzer augmented with the SpecCheck specification-based oracle, which encodes formal eBPF semantics. Detects eBPF misbehavior by comparing runtime behavior against specifications. Runs on Ubuntu 22.04; includes bug-finding, comparison, and performance evaluation tooling.

Sanidhya Kashyap

Robust Scalable Systems Software Lab

Maturity
Support
C4DT
Inactive
Lab
Active
Fuzzing
PapersArticles
Source Code

Dietary Consistency and Gut Microbiota

Analysis code for a Swiss cohort study on dietary patterns and gut microbiota diversity.

Analysis pipeline for the "Food & You" digital cohort studying associations between dietary consistency, quality metrics, and gut microbiota diversity. Includes alpha/beta diversity correlations, differential abundance analysis via BIRDMan and QIIME2, taxa correlation studies, and machine learning prediction of microbiome composition from longitudinal dietary variables.

Marcel Salathé

Salathe Lab

Maturity
Support
C4DT
Inactive
Lab
Active
FeaturesMedicalSurvey
PapersArticles
Source Code

Meal Timing Analysis

Analysis of meal timing effects on dietary quality and glucose response.

Analysis notebooks for the "Food & You" digital cohort examining meal timing and dietary quality effects on glycemic response. Implements partial correlations, linear mixed effects models, mediation analysis, weekend effects, and incremental area under the curve (iAUC) determinants using CGM-derived and dietary intake data.

Marcel Salathé

Salathe Lab

Maturity
Support
C4DT
Inactive
Lab
Active
MedicalSurveyTime Series
PapersArticles
Source Code

Food Additive Gut Microbiota

Code for analyzing food additive exposure and its impact on gut microbiota diversity.

Analysis pipeline using non-negative matrix factorization (NMF) to identify additive co-consumption patterns from MyFoodRepo barcode logs and OpenFoodFacts ingredient data, followed by z-score normalization, composite contrast metrics, and BIRDMan differential abundance analysis to assess associations between food additive mixtures and gut microbiota diversity in the "Food & You" digital cohort.

Marcel Salathé

Salathe Lab

Maturity
Support
C4DT
Inactive
Lab
Active
FeaturesMedicalSurvey
PapersArticles
Source Code

LibRRD

Build and CLI deployment tool with a web GUI component.

Build orchestration tool using SBT for compilation, with a Node.js CLI (v20+) and a statically served browser-based GUI. Deploy with `sbt deploy`; run CLI via `node cli/cli.js`; serve the GUI from the gui/ directory.

Clément Pit-Claudel

Systems and Formalisms Lab

Maturity
Support
C4DT
Inactive
Lab
Active
Development
PapersArticles
Source Code

dvar-track

Emacs Lisp tool for tracking dynamic variable dependencies with eldoc support.

Emacs Lisp tool for static analysis of dynamic variable (defvar) dependencies in Elisp and C sources. Provides dependency graphs, reporting, eldoc integration for on-the-fly documentation, and docstring annotations. Includes C analysis patches targeting Emacs internals.

Clément Pit-Claudel

Systems and Formalisms Lab

Maturity
Support
C4DT
Inactive
Lab
Active
Code Analysis
PapersArticles
Source Code

Logical Pinning

Coq formalization of logical pinning for container-internal pointers.

Coq formalization of logical pinning for container-internal (interior) pointers. Covers core libraries and case studies for linked lists and binary trees. Provides automated (make) and manual build paths with explicit dependencies. Accompanies a published research paper.

Clément Pit-Claudel

Systems and Formalisms Lab

Maturity
Support
C4DT
Inactive
Lab
Active
Formal Methods
PapersArticles
Source Code

Kyber

Cryptographic library

Kyber is a Go cryptographic library providing abstract group interfaces and concrete implementations of elliptic curves (including Ed25519 and Ristretto255), pairings, and hash-to-curve functions. It underpins threshold signatures, distributed key generation, verifiable random functions, and other advanced protocols used throughout the DEDIS ecosystem.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Active
DecentralizedEncryptionLow-Level
PapersArticles
Source Code
Contact

DRand

Publicly verifiable randomness

DRand produces publicly verifiable, unbiasable, and unpredictable random beacons using threshold BLS signatures over pairing-friendly elliptic curves. A network of independent nodes (League of Entropy) runs a distributed key generation (DKG) protocol and then periodically produces randomness that any third party can verify against the group public key.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
Byzantine ResilienceDecentralized
PapersArticles
Project Homepage
Source Code
Contact

matchertext

Matchertext is a syntactic discipline that allows strings in one compliant language to be embedded verbatim without escaping (e.g., via cut-and-paste) into itself or another compliant language.

Matchertext defines a syntactic invariant requiring that all bracket-like characters (parentheses, brackets, braces) are balanced within any well-formed string in a compliant language. This property enables verbatim, zero-escaping embedding of matchertext-compliant strings across language boundaries, eliminating injection vulnerabilities and parser ambiguity that arise from ad-hoc escaping conventions.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
Code Analysis
PapersArticles
Source Code
Contact

EPFL Combinational Benchmark Suite

Benchmarks for logical gates optimizer

A collection of combinational VHDL benchmark circuits covering arithmetic, random control, and MtM categories, designed to stress-test logic synthesis and optimization algorithms. The suite includes circuits with known best gate counts and serves as a standard reference for publications in the IWLS and CAD communities.

Giovanni De Micheli

Integrated Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
Low-Level
PapersArticles
Source Code

Igor

Crash deduplication through root-cause clustering

Fuzzing has emerged as the most effective bug-finding technique. The output of a fuzzer is a set of proof-of-concept (PoC) test cases for all observed “unique” crashes. It costs developers substantial efforts to analyze each crashing test case. This, mostly manual, process has lead to the number of reported crashes out-pacing the number of bug fixes. Igor is a tool for automated crash grouping/deduplication.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Active
Fuzzing
PapersArticles
Source Code

3D-ICE

Thermal simulator for stacked integrated circuits

3D-ICE is a compact transient thermal simulator for 3D stacked integrated circuits with inter-tier liquid cooling. It models heat dissipation using an equivalent resistor-capacitor network and solves the resulting system of differential equations to predict temperature distributions across stacked dies, supporting both linear and nonlinear thermal models with pluggable heat sink configurations.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
Low-LevelSimulator
PapersArticles
Project Homepage
Source Code

x-heep

Open-source RISC-V system-on-chip for healthcare monitoring

The X-HEEP project is an open-source RISC-V microcontroller designed for continuous healthcare monitoring. It provides a customizable microcontroller unit (MCU) composed of an open-source, industrial-grade RISC-V CPU, common peripherals, and memories, which users can extend with their own accelerators. The project targets both FPGA and ASIC implementations, with support for Xilinx-based FPGAs.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
DevelopmentInternet of ThingsMedical
PapersArticles
Project Homepage
Source Code

ByzFL

Python library for Byzantine-resilient federated learning, compatible with PyTorch and NumPy.

ByzFL implements state-of-the-art robust aggregators (Trimmed Mean, Geometric Median, Krum, Centered Clipping, etc.) and pre-aggregators (NNM, Bucketing, ARC) as well as Byzantine attack strategies (IPM, ALIE, Sign Flipping, Mimic, …) fully compatible with PyTorch tensors and NumPy arrays. It also ships a complete federated learning simulation pipeline for systematic benchmarking under adversarial conditions.

Rachid Guerraoui

Distributed Computing Lab

Maturity
Support
C4DT
Inactive
Lab
Active
Byzantine ResilienceDistributed LearningFederated LearningMachine Learning
PapersArticles
Project Homepage
Source Code

Choix

Library that provides inference algorithms for models based on Luce's choice axiom

Choix provides inference algorithms for probabilistic models based on Luce's choice axiom, including the Bradley-Terry and Plackett-Luce models. It supports pairwise comparisons, rankings, and top-k choices, with efficient optimization routines such as I-LSR and MM algorithms.

Matthias Grossglauser

Computer Communications and Applications Laboratory 4

Maturity
Support
C4DT
Inactive
Lab
Active
Low-Level
PapersArticles
Source Code
Contact

kickscore

Dynamic skill rating system

Kickscore can be used to understand & visualize the skill of players (or teams) competing in pairwise matches, and to predict outcomes of future matches. It extends the Elo rating system and TrueSkill.

Matthias Grossglauser

Computer Communications and Applications Laboratory 4

Maturity
Support
C4DT
Inactive
Lab
Active
Predictor
PapersArticles
Source Code
Contact

Stainless

Verification framework for a subset of the Scala programming language

Stainless is a tool for verifying Scala programs developed by the LARA. It can verify that your program is correct for all inputs, it can report inputs for which your program fails when they exist, and it can prove that functions do not loop.

Viktor Kunčak

Lab for Automated Reasoning and Analysis

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Active
Code AnalysisStatic Analysis
PapersArticles
Project Homepage
Source Code
Contact

Inox

Constraint solver for Stainless

Inox can be used to prove constraints on methods in a program, to attest the correctness of these parts.

Viktor Kunčak

Lab for Automated Reasoning and Analysis

Maturity
Support
C4DT
Inactive
Lab
Active
Low-Level
PapersArticles
Source Code
Contact

LISA

A Proof Assistant based on first order logic, sequent calculus and set theory.

LISA is a proof assistant built on first-order logic and sequent calculus, with a foundational layer in axiomatic set theory. It provides a kernel that checks proof certificates for validity and a higher-level tactic language in Scala for constructing and automating proofs interactively.

Viktor Kunčak

Lab for Automated Reasoning and Analysis

Maturity
Support
C4DT
Inactive
Lab
Active
Code AnalysisDevelopment
PapersArticles
Source Code
Contact

Scallion

A library for writing parsers and pretty printers for LL(1) languages in Scala.

Scallion is a combinator-based library for constructing LL(1) parsers and pretty printers in Scala, using a zipper-based parsing-with-derivatives algorithm for efficient and predictable parsing. A single grammar definition drives both directions, guaranteeing consistency between parsing and printing.

Viktor Kunčak

Lab for Automated Reasoning and Analysis

Maturity
Support
C4DT
Inactive
Lab
Active
Development
PapersArticles
Source Code

Refiner

Reasoning framework with feedback on intermediate steps

REFINER is an interaction-based framework for natural language reasoning tasks. It has a CRITIC model that provides structured feedback on intermediate reasoning steps, and a GENERATOR model that solves the reasoning task by first generating intermediate steps. The core idea is the interaction between the generator and critic, where the generator's steps are improved via feedback from the critic.

Robert West

Data Science Lab

Maturity
Support
C4DT
Inactive
Lab
Active
Machine LearningNatural LanguageOptimization
PapersArticles
Source Code

AIcrowd

Open platform for data science challenges

AIcrowd is an open crowdsourcing platform for reproducible AI research challenges. It provides tooling for dataset hosting, submission evaluation, leaderboards, and baseline implementations. The platform supports a wide variety of task types including reinforcement learning, natural language processing, and biomedical AI, and has hosted challenges in partnership with academic and industry organizations.

Marcel Salathé

Salathe Lab

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Active
CrowdSource
PapersArticles
Project Homepage
Source Code
Contact

QFlex

State-of-the-art modeling tools for the computer architecture community

QFlex is a full-system cycle-accurate computer architecture simulator of multi-node computer systems.

Babak Falsafi

Parallel Systems Architecture Lab

Maturity
Support
C4DT
Inactive
Lab
Active
Network
PapersArticles
Project Homepage
Source Code

Graphiti

Verified graph rewriting system for dataflow circuits.

Lean 4 framework for verified graph rewriting targeting dataflow circuits. Rewrites are formally proved correct in Lean; an oracle component (Rust) handles matching. The codebase is structured into a reviewed core library, per-project modules, and an experimental area. It reads and writes Dynamatic-compatible DOT graphs and emits a JSON log of every rewrite step.

Thomas Bourgeat

Verification and Computer Architecture Lab

Maturity
Support
C4DT
Inactive
Lab
Active
CompilerFormal MethodsHardware
PapersArticles
Source Code

RegElk

Linear regex engine for JavaScript in OCaml

This is a linear regular expression engine for a subset of JavaScript regexes, implemented in OCaml. It supports features like nullable quantifiers, capture reset, and lookarounds with linear time and space complexity. The engine is based on an extension of the PikeVM algorithm, described in the paper "Linear Matching of JavaScript Regular Expressions".

Clément Pit-Claudel

Systems and Formalisms Lab

Maturity
Support
C4DT
Inactive
Lab
Active
DevelopmentOptimization
PapersArticles
Source Code

Dynamatic

High-level synthesis compiler

Dynamatic is an academic, open-source high-level synthesis compiler that produces synchronous dynamically-scheduled circuits from C/C++ code. It delivers significant performance improvements compared to state-of-the-art commercial tools in specific situations.

Paolo Ienne

Processor Architecture Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
Low-Level
PapersArticles
Project Homepage
Source Code

HardCilk

Cilk-like task parallelism for FPGAs

HardCilk implements Cilk-style fork-join task parallelism directly in hardware for FPGAs, enabling dynamic task spawning and synchronization at the circuit level. It brings the expressive parallel programming model of Cilk to reconfigurable computing without a software runtime.

Paolo Ienne

Processor Architecture Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
ConcurrencyHardwareLow-Level
PapersArticles
Source Code

CoMiGS

On-device collaborative language modeling via a mixture of generalists and specialists

CoMiGS introduces a Mixture-of-Experts (MoE) architecture for personalized federated fine-tuning of language models on heterogeneous devices. Experts are categorized as generalists (whose parameters are aggregated across clients via FedAvg) or specialists (whose parameters remain local), and the split is determined through a bi-level optimization formulation. The method handles both system heterogeneity (varying LoRA ranks per device) and data heterogeneity, outperforming baselines including FedAvg, FlexLoRA, HetLoRA, FFA-LoRA, FDLoRA, and pFedMoE on standard benchmarks.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
DecentralizedFederated LearningLarge Language Model
PapersArticles
Source Code

Scala.js

Use Scala to develop Web frontends

Scala.js optimizes your Scala code into highly efficient JavaScript. Strong typing guarantees your code is free of silly mistakes. You can use any JavaScript library right from your Scala.js code, either in a statically or dynamically typed way.

Martin Odersky

Programming Methods Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
Development
PapersArticles
Project Homepage
Source Code

Scala Native

Generate native code from Scala

Native is an optimizing ahead-of-time compiler and lightweight managed runtime designed specifically for Scala. It features low-level primitices, seamless interop with native code and instant startup time.

Martin Odersky

Programming Methods Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
Development
PapersArticles
Project Homepage
Source Code

Scala 3

Scala 3 release updates and new features

Scala 3 implements many language changes and improvements over Scala 2. It strengthens Scala's foundations, makes the language easier and safer to use, and improves the consistency and expressiveness of language constructs. Key changes include intersection and union types, trait parameters, given instances, opaque type aliases, restricted implicit conversions, and a new metaprogramming model.

Martin Odersky

Programming Methods Laboratory

Maturity
Support
C4DT
Inactive
Lab
Active
Development
PapersArticles
Project Homepage
Source Code

VirTues

Foundation model for spatial proteomics enabling multi-scale tissue analysis for clinical diagnostics and biomedical discovery

VirTues is a foundation model for spatial proteomics that learns marker-aware, multi-scale representations from multiplex tissue imaging data. From a single pretrained backbone it supports zero-shot annotation across heterogeneous marker panels, cell typing, niche annotation, spatial biomarker discovery, and patient stratification.

Charlotte Bunne

Artificial Intelligence in Molecular Medicine

Maturity
Support
C4DT
Inactive
Lab
Active
Computer VisionDeep Neural NetworksImagesMachine LearningMedical
PapersArticles
Project Homepage
Source Code

DEDIS Ledger Architecture

Modular abstractions to perform Distributed Ledger (DL) operations

With Dela you can: learn the architecture of a distributed ledger, run your blockchain / distributed ledger, implement and test your new idea that will revolutionize the blockchain world by adding your new module’s implementation to the Dela ecosystem.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
Support
C4DT
Retired
Lab
Unknown
Byzantine ResilienceDecentralized
PapersArticles
Project Homepage
Source Code
Contact

Calypso

Encrypted storage with access control

Calypso implements on-chain secrets management using a distributed key generation protocol (DKG) combined with threshold decryption, so that decryption keys are never held by a single node. Access policies are expressed as Distributed Access Right Controls (DARC) stored on the OmniLedger blockchain, providing a fully auditable and decentralized enforcement mechanism.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Retired
Lab
Unknown
Access ControlConsensusEncryption
PapersArticles
Source Code
Contact

OmniLedger

Ethereum-like blockchain implementation based on skipchains

OmniLedger is a ByzCoin-based blockchain that achieves high throughput via UTXO sharding and collective signing (CoSi) for block finality. It uses skipchains for efficient client-side verification of the chain, and exposes a smart-contract model (DARC) that supports both native Go contracts and EVM-compatible Ethereum contracts.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Retired
Lab
Unknown
Byzantine Resilience
PapersArticles
Source Code
Contact

Free GAEN alternative

LibreCHovid - Alternative to GAEN for contact tracing applications

Apple and Google developed the Google-Apple Exposure Notification system in order to provide a uniform API for contact-tracing apps while preserving user privacy. On Android, GAEN is integrated in the Google Play services, which is proprietary. As a consequence, users wanting to keep a free system, as well as phone manufacturers such as Huawei who don't have Google Play services cannot use these applications. This project aims to offer an alternative framework, based on open-source code, which requires only minimal changes to the apps.

Mathias Payer

HexHive Group

Maturity
PrototypeIntermediateMature
Support
C4DT
Retired
Lab
Unknown
BluetoothMobile AppProtocol
PapersArticles
Source Code

E-ID Hands-on Workshop

A hands-on workshop showing how cryptography can preserve privacy and security of an E-ID solution.

To protect the privacy of the users of E-ID systems, several technical measures need to be taken. This workshop demonstrates the workings of Self-Sovereign Identities, using selective disclosure for data protection, and explores further safeguards like unlinkability through BBS+ and zero-knowledge proofs to share only essential information.

Jean-Pierre Hubaux

Center For Digital Trust

Maturity
Support
C4DT
Retired
Lab
Unknown
Access ControlDevelopment
PapersArticles
Source Code

Garfield

System support for byzantine machine learning

Garfield is a library to build Byzantine machine learning (ML) applications on top of popular frameworks such as TensorFlow and PyTorch. We show how to use Garfield to build different architectures for ML applications like single server, multiple workers (SSMW), multiple servers, multiple workers (MSMW), and fully decentralized architecture.

Rachid Guerraoui

Distributed Computing Lab

Maturity
PrototypeIntermediateMature
Support
C4DT
Retired
Lab
Unknown
Byzantine ResilienceDecentralizedDistributed LearningFederated LearningPyTorchTensorFlow
PapersArticles
Source Code

AT2

Asset Transfer System using lightweight broadcast-based primitives instead of consensus

We replace consensus with a lightweight building block -- a broadcast-based primitive -- to obtain a full-fledged asset transfer system (cryptocurrency). This system is more efficient and more scalable compared with Bitcoin and consensus-based solutions. AT2 is the first system to bypass consensus algorithms for implementing asset transfers.

Rachid Guerraoui

Distributed Computing Lab

Maturity
PrototypeIntermediateMature
Support
C4DT
Retired
Lab
Unknown
Byzantine ResilienceConsensusDecentralized
PapersArticles
Source Code

Stainless for smart contracts

Stainless fork aimed at smart contracts

Stainless is a tool for verifying Scala programs developed by the LARA. This fork can perform verification of Smart Contracts.

Viktor Kunčak

Lab for Automated Reasoning and Analysis

Maturity
PrototypeIntermediateMature
Support
C4DT
Retired
Lab
Unknown
Code AnalysisStatic Analysis
PapersArticles
Project Homepage
Source Code
Contact

MedCo

Privacy-preserving cohort exploration

MedCo is a federated system for privacy-preserving clinical cohort exploration across multiple hospital sites. It combines homomorphic encryption, secure multi-party computation, and differential privacy to allow researchers to run aggregate queries on sensitive patient data without any site exposing individual records.

Jean-Pierre Hubaux

Laboratory for Data Security

Maturity
PrototypeIntermediateMature
Support
C4DT
Retired
Lab
Unknown
Homomorphic EncryptionMedicalSecure Multi-Party Computation
PapersArticles
Project Homepage
Source Code

Decentralized Privacy-Preserving Proximity Tracing

A privacy preserving exposure notification protocol to inform people about possible exposure to the COVID-19 virus

The goal of the DP3T protocol is to simplify and accelerate the process of identifying people who have been in contact with a positively CoVid-19 tested person, thus providing a technological foundation to help slow the spread of the SARS-CoV-2 virus. The system aims to minimise privacy and security risks for individuals and communities and guarantee the highest level of data protection.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Retired
Lab
Unknown
BluetoothProtocol
PapersArticles
Source Code

CrowdNotifier

Decentralized Privacy-Preserving Presence Tracing

CrowdNotifier is a protocol that can complement manual contact tracing by efficiently notifying visitors of venues and events with SARS-CoV-2-positive attendees. It provides strong privacy and abuse- resistance, and show that it can scale to handle notification at a national scale.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Retired
Lab
Unknown
Protocol
PapersArticles
Source Code

lightarti-rest

A Rust library to perform REST requests over the Tor network for mobile applications.

This library builds upon the arti-library. It wraps the arti-library for easy use by mobile applications. Instead of downloading the whole dictionary of the Tor-nodes every hour, a prepared dictionary is downloaded once per week. This reduces bandwidth requirements, as the whole dictionary is around 3MB. The prepared dictionary only weighs 200kB.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
Support
C4DT
Retired
Lab
Unknown
Mobile AppTor
PapersArticles
Source Code

MLBench

Benchmarking of distributed ML

Framework for distributed machine learning. Its purpose is to improve transparency, reproducibility, robustness, and to provide fair performance measures as well as reference implementations, helping adoption of distributed machine learning methods both in industry and in the academic community. Besides algorithm comparison, a main use case is to help the selection of hardware (CPU, GPU) used to run AI applications, as well as how to connect it into a cluster to get a good cost/performance tradeoff.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Retired
Lab
Unknown
Benchmark
PapersArticles
Project Homepage
Source Code

Drynx

Decentralized, secure, verifiable system for statistical queries and machine learning on distributed datasets

Drynx allows to create privacy-preserving queries on encrypted datasets that are stored at different data providers who don't want to share the original data. Different types of statistical queries are possible, like average, standard deviation, linear and logistic regressions - all using homomorphic encryption, which means that the data is never shared in cleartext.

Jean-Pierre Hubaux

Laboratory for Data Security

Maturity
PrototypeIntermediateMature
Support
C4DT
Retired
Lab
Unknown
Homomorphic EncryptionSecure Multi-Party Computation
PapersArticles

SPINDLE

Scalable, privacy-preserving, distributed learning on distributed datasets

SPINDLE allows to train and evaluate generalized linear models on datasets that are stored at different data providers who don't want to share the original data.

Jean-Pierre Hubaux

Laboratory for Data Security

Maturity
Support
C4DT
Retired
Lab
Unknown
Homomorphic EncryptionSecure Multi-Party Computation
PapersArticles

Fine-tuning pitfalls

Hands-on workshop on LLM fine-tuning pitfalls

A practical workshop covering common failure modes in LLM fine-tuning, including issues with dataset quality, learning rate schedules, overfitting, catastrophic forgetting, and evaluation methodology. Participants work through concrete examples using open-source tooling to identify and correct these problems in realistic fine-tuning scenarios.

Volkan Cevher

Laboratory for Information and Inference Systems

Maturity
Support
C4DT
Retired
Lab
Unknown
FailureMachine Learning
PapersArticles

HEEPstor

Framework for deploying post-training quantized PyTorch models on X-HEEP RISC-V SoCs.

Open-hardware framework for post-training quantization and deployment of PyTorch models on X-HEEP RISC-V SoCs. Provides automated INT8/FP32 hybrid quantization, FPGA synthesis, a configurable systolic array, performance analysis, and debugging tools. Targets edge AI workloads.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
HardwareInternet of ThingsPyTorchRISC-V
PapersArticles
Source Code

JSONSchemaBench

Benchmark of ~10,000 real-world JSON schemas for evaluating structured output generation.

Benchmark of ~10,000 real-world JSON schemas for evaluating LLM structured output generation across efficiency, coverage, and quality dimensions. Integrates with Hugging Face datasets; supports custom engine integration and leaderboard submission.

Robert West

Data Science Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
BenchmarkLarge Language Model
PapersArticles
Source Code

Sourcerer

LLVM-based C++ type checking tool with performance overhead and fuzzing evaluation.

LLVM-based instrumentation tool enforcing C++ type safety at runtime to detect type confusion and memory safety violations. Artifact includes scripts for compatibility testing, SPEC CPU performance overhead measurement, ablation studies, and fuzzing evaluation. Docker-based setup with hardware/software requirements.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
RobustnessStatic Analysis
PapersArticles
Source Code

Pedesite App

Wear OS application for wearable sensor data acquisition, storage, and automated upload.

Wear OS application for continuous wearable sensor data acquisition, local storage, and automated batch upload to Kafka while charging. Includes ADB installation instructions, watch-face activation, Python scripts for sensor file reading, and data visualization tools.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Internet of ThingsMobile AppTime SeriesWearable
PapersArticles
Source Code

Double-Edged Shield

Study and attacks on fingerprinting users through customized ad-blocker filter lists.

Fingerprinting study using customized ad-blocker filter lists as an identification vector. Presents three attacks achieving 84% list identification, 0.72-bit entropy fingerprints, and median anonymity sets of 48 users. Provides datasets, reproducibility artifacts, and a honeypage for data collection.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
AttackFingerprinting
PapersArticles
Source Code

ICCWD

Scripts for the Image-Caption in the Wild Dataset and minor-detection method evaluation.

Scripts for downloading and evaluating the Image-Caption in the Wild Dataset (ICCWD) using three minor-detection pipelines: DeepSeek caption labeling, Amazon Rekognition image analysis, and a combined approach. Outputs TPR/FPR metrics and ROC curves.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Images
PapersArticles
Source Code

LipsLev

PyTorch implementation of certified robustness against bounded Levenshtein distance edits.

PyTorch implementation of certified robustness against bounded Levenshtein (edit) distance perturbations from the ICLR 2025 paper. Provides Lipschitz-constrained text classifiers with formal robustness certificates, training scripts, verification methods, and pre-trained models. Also covers the related Charmer method.

Volkan Cevher

Laboratory for Information and Inference Systems

Maturity
Support
C4DT
Inactive
Lab
Unknown
Natural LanguagePyTorchRobustness
PapersArticles
Source Code

LION

Framework converting linear-attention Transformers into bidirectional RNN equivalents.

Framework converting linear-attention Transformers to bidirectional RNN equivalents for fast training and efficient inference. Provides three variants (LION-Lit, LION-D, LION-S) for image classification and masked language modeling, with configurable size, masking, patch order, and format. Achieves competitive accuracy with reduced training time.

Volkan Cevher

Laboratory for Information and Inference Systems

Maturity
Support
C4DT
Inactive
Lab
Unknown
Deep Neural Networks
PapersArticles
Source Code

COINs

Community-informed graph embeddings for fast knowledge-graph link prediction.

Community-informed graph embeddings for knowledge-graph link prediction and conjunctive query answering. Two-stage prediction: community membership prediction followed by within-community node lookup. Achieves ~97% baseline accuracy at ~7.5% of original compute. ICT Innovations 2025 paper.

Volkan Cevher

Laboratory for Information and Inference Systems

Maturity
Support
C4DT
Inactive
Lab
Unknown
Deep Neural Networks
PapersArticles
Source Code

FineWeb2-HQ

Multilingual data filtering tools for LLM pretraining using FastText and Transformer-MLP classifiers.

Toolset for multilingual data quality filtering using FastText language identification and Transformer-MLP quality classifiers. Supports embedding generation, model training, and automated dataset curation for the FineWeb2-HQ corpus covering 20 languages. Accompanies an arXiv preprint on model-based data selection.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Large Language ModelNatural Language
PapersArticles
Source Code

MendelFuzz

Improved deterministic stage for AFL++ fuzzer, now default in AFL++ 4.10c.

Enhanced deterministic mutation stage for AFL++ improving coverage and bug-finding over the standard havoc and deterministic strategies. Merged as the default stage in AFL++ 4.10c. Artifact includes reproducibility scripts for MAGMA and FuzzBench evaluations with documented data quirks.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
Fuzzing
PapersArticles
Source Code

GSR Monitor Board

Hardware validation board for HEEPidermis SoC with low-noise power regulation and GSR measurement.

Hardware validation board for the HEEPidermis SoC. Features low-noise power regulation, 16-bit current DAC (40 nA–10 µA), precision resistors, level-crossing and 12-bit ΣΔ ADCs, I²C interface (1.65–5.5 V), and Arduino library support. Includes schematics and WEMAC dataset integration.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
HardwareInternet of ThingsWearable
PapersArticles
Source Code

PRAU

Open-source configurable Posit arithmetic unit supporting extended numerical operations.

Open-source configurable hardware Posit Arithmetic Unit supporting posit sizes from 8–64 bits. Implements standard posit operations plus optional log-approximate multiplication, division, sqrt, and quire (exact accumulation). Derived from the PAU in the PERCIVAL RISC-V core; used by Coprosit and PHEE projects.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
HardwareLow-LevelRISC-V
PapersArticles
Source Code

Coprosit

Posit and quire arithmetic coprocessor compliant with the RISC-V eXtension interface.

Posit and quire arithmetic coprocessor compliant with the RISC-V CV-X-IF eXtension interface, wrapping the PRAU execution unit. Configurable posit sizes (8–64 bits) and optional operations (log-approximate multiply, divide, sqrt, quire). Integrated into X-HEEP and PHEE; accessible via make configuration flags.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
HardwareLow-LevelRISC-V
PapersArticles
Source Code

CV-X-IF Coprocessor

Sample RISC-V coprocessor implementing a custom BITREV instruction via the CV-X-IF interface.

Sample RISC-V coprocessor implementing the R-type BITREV instruction via the CV-X-IF interface for the cv32e40px core. Includes hardware decode and execute stages and a C usage example. Serves as a template for custom instruction extension development.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
HardwareLow-LevelRISC-V
PapersArticles
Source Code

Minimum Days Estimation

Code for estimating the optimal study duration for reliable dietary intake information.

Analysis notebooks applying Linear Mixed Models (LMM), Coefficient of Variation (CV), and Intraclass Correlation Coefficient (ICC) methods to determine the minimum dietary monitoring duration for reliable intake estimation. Analyzes data from the "Food & You" digital cohort using Python scientific computing libraries (pandas, numpy, statsmodels, pingouin).

Marcel Salathé

Salathe Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
PredictorSurvey
PapersArticles
Source Code

MyFoodRepo

Nutrition tracking platform with food image segmentation and annotation tools.

Ruby on Rails web platform implementing food image segmentation via an AI/CrowdAI API, product database synchronization with FoodRepo, REST API, and annotation management workflows. Supports direct upload, segmentation status tracking, and data serialization for large-scale digital dietary cohort studies.

Marcel Salathé

Salathe Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
ImagesMobile App
PapersArticles
Source Code

Proof-of-Personhood

Proof that you are an actual person while keeping your anonymity

Fake accounts are a big problem on the internet, while not all people want or can identify themselves using state-issued identities. Proof-of-personhood proposes to use personhood parties where people have to physically show up and get identified as real persons. Using cryptographic primitives, it is then possible to prove that you were part of that group, without revealing your individual identity.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Anonymity
PapersArticles
Source Code
Contact

Odyssey

Auditable Sharing and Management of Sensitive Data Across Jurisdictions

Odyssey is a set of applications and tools that enables sharing of sensitive data between multiple distrustful parties. This project uses state-of-the-art secret management service on the blockchain coupled with an enclave delivery mechanism.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Access ControlEncryption
PapersArticles
Project Homepage
Source Code
Contact

Onet

Library for simulation and deployment of decentralized, distributed protocols

Onet is a Go framework for building overlay-network protocols and distributed applications. It provides abstractions for trees of nodes, protocol dispatch, service registration, and message routing over TLS. Its built-in simulation framework allows running multi-node experiments locally or on remote machines to benchmark latency and throughput.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Low-LevelNetworkProtocolSimulator
PapersArticles
Source Code
Contact

E-voting

Decentralized, encrypted, fully anonymous e-voting

Our e-voting system is inspired by the first version of Helios where the encrypted ballots are shuffled and anonymized before they are decrypted. Instead of the shuffle used in Helios, we implemented a Neff shuffle which is much faster than the original Helios shuffle.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Anonymity
PapersArticles
Source Code
Contact

PURBs

Metadata leakage prevention for encrypted communication and storage

PURBs (Padded Uniform Random Blobs) encode encrypted messages so that ciphertexts reveal no metadata — such as the number or type of recipients — to passive observers. The encoding uses Elligator-based public keys embedded at uniformly random positions, and pads ciphertexts to a size drawn from a geometric distribution to hide payload length.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Encryption
PapersArticles
Source Code
Contact

Threshold Logical Clocks

Solving the consensus problem in a new way by creating a global clock

An attempt to simplify asynchronous consensus by building atop a novel threshold logical clock abstraction, which enables upper layers to operate as if on a synchronous network. This approach yields an asynchronous consensus protocol for fail-stop nodes that may be simpler and more robust than Paxos and its leader-based variants, requiring no common coins and achieving consensus in a constant expected number of rounds.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Byzantine ResilienceConsensus
PapersArticles
Source Code
Contact

SimNet

SimNet is a tool to simulate decentralized applications in the cloud

SimNet provides multiple configurations to affect the topology so that specific link between two nodes can have a delay or a loss of packets for instance.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Simulator
PapersArticles
Source Code
Contact

HBT

Token for conflicts zone

HBT implements an anonymous token system on a distributed ledger targeting high-risk, low-connectivity environments. It uses cryptographic anonymity techniques to ensure transaction unlinkability while maintaining auditability for authorized parties, and is designed to operate over intermittently connected networks.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
AnonymityDecentralized
PapersArticles
Project Homepage
Source Code
Contact

determ-checker

Find non-determinism in golang code

determ-checker is a static analysis tool for Go programs that detects sources of non-determinism such as map iteration order, goroutine scheduling races, and randomized data structure traversal. It is particularly relevant for consensus and blockchain code where non-deterministic execution can cause protocol forks.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Code Analysis
PapersArticles
Source Code
Contact

APIR

Authenticated Private Information Retrieval

APIR extends classical Private Information Retrieval (PIR) with cryptographic authentication guarantees. It provides both single-server and multi-server protocols, combining homomorphic-encryption-based PIR queries with Merkle-tree commitments or linear sketch-based MACs so that a client can detect a dishonest server while revealing nothing about the queried index.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
Access ControlAnonymity
PapersArticles
Source Code
Contact

QuePaxa

Asynchronous consensus algorithm that escapes the tyranny of timeouts

QuePaxa is a crash fault-tolerant asynchronous consensus algorithm that improves upon Multi-Paxos, Raft, Rabia, and EPaxos. Its three main innovations are: an asynchronous consensus core with a one-round-trip fast path, hedging-delay instead of traditional timeouts, and dynamic runtime protocol tuning. The implementation is written in Go and includes integration tests and AWS deployment scripts for artifact evaluation against competing consensus algorithms.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
Byzantine ResilienceConsensusNetworkProtocol
PapersArticles
Source Code
Contact

Limix

Geo-distributed database with overlapping jurisdictional zone isolation

Limix is a geo-distributed database built on CockroachDB, organized into overlapping jurisdictional zones. Each physical node can run multiple CockroachDB instances — one per zone it belongs to. Coordination uses DEDIS onet v2 for intra-zone membership dissemination and simnet for container orchestration. Clients perform strongly-consistent per-key reads and writes across relevant zones. Experiments evaluate partition tolerance and latency on random synthetic, CAIDA Ark, and real AWS topologies.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
DecentralizedNetwork
PapersArticles
Source Code
Contact

Memoro

Highly detailed heap profiler

Modern software systems heavily use the memory heap. As systems grow more complex and compute with increasing amounts of data, it can be difficult for developers to understand how their programs actually use the bytes that they allocate on the heap and whether improvements are possible. Memoro as built to answer this question of heap usage efficiency. It uses a combination of static instrumentation, subroutine interception, and runtime data collection to build a clear picture of exactly when and where a program performs heap allocation, and crucially how it actually uses that memory. Memoro also introduces a new visualization application that can distill collected data into scores and visual cues that allow developers to quickly pinpoint and eliminate inefficient heap usage in their software.

James Larus

Very Large Scale Computing Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
DevelopmentLow-Level
PapersArticles
Source Code

Abacus

Precise side-channel analysis

Side-channel attacks allow adversaries to infer sensitive information from non-functional characteristics. Typical side-channel detection tools are able to identify many potential vulnerabilities, but do not provide information regarding their severity, making it difficult for developers to decide whether to address them, as their leaks are often negligible in practice. Abacus is able to precisely quantify the leaked information and finds flaws that would otherwise be difficult for a developer to find.

James Larus

Very Large Scale Computing Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
DevelopmentProtection
PapersArticles
Source Code

Bitfiltrator

Bitfiltrator is an automated bitstream parameter extraction tool for Xilinx UltraScale and UltraScale+ FPGAs.

Prior work has reverse-engineered parts of the bitstream format for security or debugging/instrumentation activities, but no paper has explained how to do this reverse engineering systematically! Our work bridges this gap by explaining: (1) the various parameters needed to navigate a bitstream correctly, (2) the experiments to obtain them, and (3) the many pitfalls and erroneous assumptions to avoid while undertaking this endeavor. We demonstrate our technique by using it to extract the bitstream format of initial LUT equations, LUTRAM contents, BRAM contents, and register values in Xilinx UltraScale and UltraScale+ FPGAs. Our methods are implemented in an open-source tool, Bitfiltrator [1], that can extract device layouts and architecture-specific bitstream formats for these cells automatically and without physical access to an FPGA.

James Larus

Very Large Scale Computing Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
DebugLow-Level
PapersArticles
Source Code

QualityCrowd

Crowdsourcing image quality assessment

Video quality assessment with subjective testing is both time consuming and expensive. An interesting new approach to traditional testing is the so-called crowdsourcing, moving the testing effort into the internet. QualityCrowd allows to perform subjective quality assessment with crowdsourcing. QualityCrowd allows codec independent quality assessment with a simple web interface, usable with common web browsers.

Touradj Ebrahimi

Multimedia Signal Processing Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
Images
PapersArticles
Source Code

Point cloud web renderer

Show point cloud files using three.js

The renderer supports visualization of point clouds with real-time interaction, while viewing conditions can be configured. The user is able to choose between either an adaptive, or a fixed splat size rendering mode in order to display the models. The renderer supports both PLY and PCD point cloud file formats. The current settings have been optimized for voxelized contents, without this limiting its usage, since any point cloud can be displayed independently of its geometric structure.

Touradj Ebrahimi

Multimedia Signal Processing Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
Images
PapersArticles
Source Code

Learned Residual PCC

Learning residual coding for point clouds

Recent advancements in acquisition of three-dimensional models have been increasingly drawing attention toimaging modalities based on the plenoptic representations, such as light fields and point clouds. Since point cloudmodels can often contain millions of points, each including both geometric positions and associated attributes,efficient compression schemes are needed to enable transmission and storage of this type of media. This project presents a detachable learning-based residual module for point cloud compression that allows for efficientscalable coding.

Touradj Ebrahimi

Multimedia Signal Processing Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
Images
PapersArticles
Source Code
Contact

Error Resilient Recurring Gallery Building

Correctly re-identify people in different images

In person re-identification, people must to be correctly identified in images that come from different cameras or are captured at different points in time. In the open-set case, the above needs be achieved for persons that have not been previously observed. We propose a universal method for building a multi-shot gallery of observed reference identities recurrently online.

Touradj Ebrahimi

Multimedia Signal Processing Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
Computer Vision
PapersArticles
Source Code
Contact

Mesh sampling

Generate point cloud based on mesh

It is composed of two scripts that perform mesh sampling. They were developed with the goal of generating point cloud datasets for training of learning-based compression algorithms from large-scale mesh datasets such as ModelNet and ShapeNet.

Touradj Ebrahimi

Multimedia Signal Processing Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
Computer Vision
PapersArticles
Source Code

Point-cloud compressor

Learning-based compression of point cloud geometry and color

It includes scripts for training and testing a neural network architecture for compression of geometry and color of point clouds. The compression model is based on a convolutional auto-encoder architecture that operates on block partitions of a point cloud.

Touradj Ebrahimi

Multimedia Signal Processing Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
Computer Vision
PapersArticles
Source Code

ProCam

Privacy-preserving camera system tracking individual body temperatures

Many contagious diseases such as Covid19 can induce high temperatures and fever in a significant number of affected individuals. Contact tracing of such individuals and analysis of their behaviour and interactions with others and their environment can be a useful tool to contain spread of contagious diseases. A key component of ProCam is an innovative camera that combines infrared/thermal and visible light sensors while securing the visible light images. Privacy-preserving camera systems are important as more and more people are concerned about the ethics of massive surveillance and invasion of privacy. ProCam solves this by using a cryptographic tool to hide the visible light images, only allowing to retrieve identites with the use of cryptographic keys.

Touradj Ebrahimi

Multimedia Signal Processing Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
AnonymityImagesMedical
PapersArticles
Project Homepage
Source Code

AdMiRe EPFL

EPFL's contribution to the AdMiRe project: the Background Extraction and Super Resolution modules, with input and output as NDI streams.

Currently, TV audiences can only interact with the shows they are watching through social media or hybrid broadcast broadband TV. Additionally, content creators in film, TV or live productions face technical challenges that result in less engaging content. The EU-funded AdMiRe project aims to tackle both these problems by developing, validating and demonstrating innovative solutions based on mixed reality technology. These solutions will enable audiences at home to be incorporated into the live TV programme they are watching and to interact with people in the TV studio. They will also provide content creators with tools that radically improve talent immersion and interaction with computer-generated elements.

Touradj Ebrahimi

Multimedia Signal Processing Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
Images
PapersArticles
Project Homepage
Source Code

Quick ADC

Fast distance computation for searches in large databases

Quick ADC is an implementation of fast distance computation techniques for nearest neighbor search in large databases of high-dimensional vectors.

Anne-Marie Kermarrec

Scalable Computing Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Database
PapersArticles
Source Code

ProteusTM

Automatic transactional memory adaptation

There are many designs and implementations of Transactional Memory, and that is not a coincidence: different ones perform best on different workloads. One cannot expect the developer to be aware of such concerns, as it is even contradictory with the simplicity advocated by the use of the transactional abstraction. With ProteusTM, the developer writes transactions and the underlying implementation uses techniques to automatically change Transactional Memory algorithms, parallelism degree, and others.

Anne-Marie Kermarrec

Scalable Computing Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
ConcurrencyDevelopment
PapersArticles
Source Code

EPFL Logic Synthesis Libraries

Collection of modular libraries for the development of logic synthesis applications

A suite of modular C++ libraries supporting logic synthesis research, including data structures for Boolean networks, And-Inverter Graphs (AIGs), and k-LUT networks, along with algorithms for technology mapping, rewriting, and simulation. Designed for composability and used as the foundation of the ABC-compatible EPFL synthesis flow.

Giovanni De Micheli

Integrated Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Low-Level
PapersArticles
Source Code

CirKit

Logic synthesis and optimization framework

CirKit is a C++ framework for logic synthesis and optimization, offering data structures for AIGs, MIGs, and other graph representations alongside algorithms for rewriting, mapping, and verification. It exposes an interactive shell and Python/shell scripting interface, and integrates with external tools such as ABC and SAT solvers.

Giovanni De Micheli

Integrated Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Low-Level
PapersArticles
Source Code

FuZZan

Faster address sanitization to improve fuzzing speed

Our evaluation shows that on the Google fuzzer test suite, FuZZan improves fuzzing throughput over ASan by 48% starting with Google’s provided seeds (52% when starting with empty seeds on the same applications). Due to this improved throughput, FuZZan discovers 13% more unique paths given the same 24 hours and finds bugs 42% faster

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
Fuzzing
PapersArticles
Source Code

USBFuzz

Virtual environment for fuzzing USB peripherals

Fuzzing USB peripherals for Linux, Android, MacOS, Windows, FreeBSD Found lots of 0days in modern kernels, allowing arbitrary compromise

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
Fuzzing
PapersArticles
Source Code

Magma

A ground-truth fuzzing benchmark suite based on real programs with real bugs

Magma is a collection of open-source libraries with widespread usage and a long history of security-critical bugs and vulnerabilities. In light of the need for better fuzzer evaluation, we front-ported bugs from previous bug reports to the latest versions of these libraries. For each ported bug, we added in-line (source-code-level) instrumentation to collect ground-truth information about bugs reached (buggy code executed) and triggered (fault condition satisfied by input). This instrumentation allows a monitoring utility to measure fuzzer progress in real time.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
BenchmarkFuzzing
PapersArticles
Project Homepage
Source Code

T-Fuzz

Fuzzing tool and crash analyzer

T-Fuzz (transformational fuzzing) improves fuzzing by detecting whenever the fuzzer is stuck and then modifying the underlying program to reach new program locations. This allows fuzzing to discover and test hard to reach code.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
DebugFailureFuzzing
PapersArticles
Source Code

BOPC

Block-oriented programming compiler

BOPC is a compiler that automatically synthesizes data-only exploits using Block Oriented Programming (BOP). It searches a program's existing code for functional blocks (SPloits) that can be chained together to express arbitrary computation, bypassing control-flow integrity (CFI) defenses without injecting new code.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
AttackExploitFuzzing
PapersArticles
Source Code

CFIXX

Protection for C++ internal code pointers

CFIXX is a runtime mitigation that protects C++ programs from attacks that overwrite internal data structures containing code pointers.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
ExploitFuzzing
PapersArticles
Source Code

HexType

Efficient detection of type confusion errors for C++

HexType is a sanitizer for C++ cast safety. The primary use case is during testing to detect type confusion issues in the code. Additionally, HexType may be run in a fuzzing environment to detect type violations.

Mathias Payer

HexHive Group

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
Fuzzing
PapersArticles
Source Code

RetroWrite

Static instrumentation of binaries for fuzzing and sanitation

RetroWrite can analyze the security of closed source binaries by applying an efficient fuzzing technique that usually only works when the source code is available. This makes it possible to trigger more errors than the default binary fuzzers available, while being much faster in finding these errors.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
BinaryFuzzing
PapersArticles
Source Code

ShadowStack

LLVM Implementation of different ShadowStack schemes for x86_64

This project provides an LLVM-based implementation of multiple shadow stack schemes for x86_64, including parallel shadow stack, shstk (Intel CET), and randomized schemes. It serves as a systematic evaluation framework for comparing the security guarantees, performance overhead, and compatibility trade-offs of different shadow stack approaches on modern hardware.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
BinaryProtection
PapersArticles
Source Code

HexPADS

Host-based, performance-counter-based attack detection system

Attacks change the execution behavior of a system. Our attack detection system HexPADS detects attacks through divergences from normal behavior using attack signatures. HexPADS collects information from the operating system on runtime performance metrics with measurements from hardware performance counters for individual processes. Cache behavior is a strong indicator of ongoing attacks like rowhammer, side channels, covert channels, or CAIN attacks. Collecting performance metrics across all running processes allows the correlation and detection of these attacks. In addition, HexPADS can mitigate the attacks or significantly reduce their effectiveness with negligible overhead to benign processes.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
Protection
PapersArticles
Source Code

FuzzGen

Automatic fuzzer generation

FuzzGen is a tool for automatically synthesizing fuzzers for complex libraries in a given environment. From a given library, it allow for creating a dedicated fuzzer, aiming at good code coverage, without human interaction.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
DebugFailureFuzzing
PapersArticles
Source Code

DataShield

Configurable data confidentiality and integrity

DataShield can, based on C/C++ languages annotation, protect unwanted access to secret data. By securing only theses datastructures, performance overhead is greatly reduced.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
DebugFailureFuzzing
PapersArticles
Source Code

HALucinator

Firmware re-hosting through abstraction layer emulation

HALucinator allow for running a firmware outside of its physical implementation. When extracting it, one can recreate the same environment as the one used by the firmware's developer thus allowing for faster exploitation.

Mathias Payer

HexHive Group

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
BinaryDebugExploit
PapersArticles
Source Code

BenchIoT

Security benchmarks for the Internet of Things

BenchIoT enables automatic evaluation of 14 metrics covering security, performance, memory usage, and energy consumption. Having a sound benchmark focused on IoT allow for robust and comparable metrics.

Mathias Payer

HexHive Group

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
BenchmarkInternet of Things
PapersArticles
Source Code

LightBlue

Automatic profile-aware debloating of bluetooth stacks

The Bluetooth standard is complex, many commands are defined but aren't needed in many use cases. It allows for a big surface attack which can be cumbersome to correctly reduce. Based on Bluetooth usage profiling, LightBlue allows to reduce the number of exposed commands understood by the firmware and the runtime, thus greatly reducing the potential future attacks that might arise.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
AttackBluetooth
PapersArticles
Source Code

dis-cover

Disassemble binaries and recover as much info as possible

dis-cover is a binary analysis tool that disassembles compiled executables and applies heuristics to recover high-level information such as function boundaries, symbols, and data structures. It aids reverse engineering and vulnerability research by reconstructing semantic information from stripped binaries.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
BinaryLow-Level
PapersArticles
Source Code

Gramatron

Grammar-aware fuzzing

Coverage-guided fuzzer that uses grammar automatons to perform grammar-aware fuzzing.

Mathias Payer

HexHive Group

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
Fuzzing
PapersArticles
Source Code

datAFLow

Data-flow based fuzzing

DatAFLow is a fuzzer built on top of AFL++. However, instead of a control-flow-based feedback mechanism, it uses a data-flow-based feedback mechanism; specifically, data flows based on def-use associations.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
Fuzzing
PapersArticles
Source Code

Midas

Systematic kernel TOCTTOU protection

During a syscall, the kernel might read multiple times a value from userspace. As theses are controlled by the unpriviliged process, it may introduce subtle changes of behavior and possible hard-to-detect attacks, a classic time-of-check-to-time-of-use bug. Midas disable this attack surface by always returning the same values during a given syscall.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
BinaryDebugProtection
PapersArticles
Source Code

GLeeFuzz

Fuzzing WebGL Through Error Message Guided Mutation

Our fuzzing technique, GLeeFuzz, guides input mutation by error messages instead of code coverage. Our key observation is that browsers emit meaningful error messages to aid developers in debugging their WebGL programs. Error messages indicate which part of the input fails (e.g., incomplete arguments, invalid arguments, or unsatisfied dependencies between API calls). Leveraging error messages as feedback, the fuzzer effectively expands coverage by focusing mutation on erroneous parts of the input. We analyze Chrome’s WebGL implementation to identify the dependencies between error-emitting statements and rejected parts of the input, and use this information to guide input mutation. We evaluate our GLeeFuzz prototype on Chrome, Firefox, and Safari on diverse desktop and mobile OSes. We discovered 7 vulnerabilities, 4 in Chrome, 2 in Safari, and 1 in Firefox. The Chrome vulnerabilities allow a remote attacker to freeze the GPU and possibly execute remote code at the browser privilege.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
Fuzzing
PapersArticles
Source Code

Evocatio

Evocatio is a bug analyzer built on top of AFL++ and AddressSanitizer. It automatically discovers a bug’s capabilities.

Evocatio automatically discovers a bug’s capabilities: analyzing a crashing test case (i.e., an input exposing a bug) to understand the full extent of how an attacker can exploit a bug. Evocatio leverages a capability-guided fuzzer to efficiently uncover new bug capabilities (rather than only generating a single crashing test case for a given bug, as a traditional greybox fuzzer does).

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
Fuzzing
PapersArticles
Source Code

WarpAttack

Bypassing CFI through Compiler-Introduced Double-Fetches

WarpAttack is a new attack vector that exploits compiler-introduced double-fetch optimizations to mount TOCTTOU attacks and bypass code-reuse mitigations.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
Attack
PapersArticles
Source Code

ViDeZZo

Dependency-aware Virtual Device Fuzzing Framework

ViDeZZo is a virtual device fuzzing framework considering both intra- and inter-message dependencies to balance fuzzing scalability and efficiency.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
Fuzzing
PapersArticles
Source Code

SieveFuzz

Optimizing Directed Fuzzing via Target-tailored Program State Restriction

We implement tripwiring-directed fuzzing as a prototype, Sieve-Fuzz, and evaluate it alongside the state-of-the-art directed fuzzers AFLGo, BEACON and the leading undirected fuzzer AFL++. Overall, across nine benchmarks, SieveFuzz’s tripwiring enables it to trigger bugs on an average 47% more consistently and 117% faster than AFLGo, BEACON and AFL++.

Mathias Payer

HexHive Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
Fuzzing
PapersArticles
Source Code

SciLens

Helper for science news evaluation

A framework helping non-experts to evaluate the quality of scientific news articles, based on indicators automatically generated using machine learning.

Karl Aberer

Distributed Information Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Natural Language
PapersArticles
Project Homepage
Source Code
Contact

Media Observatory

News landscape monitoring

A model of the media landscape that maps news channels based on their selection of subjects. The observatory has the following key objectives: giving the public access to algorithmic solutions to monitor the news landscape, developing transparent quality and credibility indicators and creating a meeting point to foster multidisciplinary research opportunities

Karl Aberer

Distributed Information Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Natural Language
PapersArticles
Project Homepage
Source Code

Mozuma

Library of ML standardised models

Mozuma is a library containing a collection of machine learning models with standardised interface to run inference, train and manage model state files.

Karl Aberer

Distributed Information Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
PyTorch
PapersArticles
Source Code

Issuer & Verifier for the Swiss e-ID 2023 sandbox

Demo for the swiss E-ID ledger. It includes an issuer and a verifier.

The issuer is a web application that allows to create a new identity and issue a credential. The verifier is a web application that allows to verify the credential.

Jean-Pierre Hubaux

Center For Digital Trust

Maturity
Support
C4DT
Inactive
Lab
Unknown
Access ControlDevelopment
PapersArticles
Source Code

DeepFool

Simple algorithm to find the minimum adversarial perturbations in deep networks

State-of-the-art deep neural networks have achieved impressive results on many image classification tasks. However, these same architectures have been shown to be unstable to small, well sought, perturbations of the images. Despite the importance of this phenomenon, no effective methods have been proposed to accurately compute the robustness of state-of-the-art deep classifiers to such perturbations on large-scale datasets. DeepFool proposes to efficiently compute perturbations that fool deep networks, and thus reliably quantify the robustness of these classifiers.

Pascal Frossard

Signal Processing Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Deep Neural Networks
PapersArticles
Source Code

ManiFool

Algorithm for evaluating the invariance properties of deep networks

Deep convolutional neural networks have been shown to be vulnerable to arbitrary geometric transformations. However, there is no systematic method to measure the invariance properties of deep networks to such transformations. ManiFool is a simple yet scalable algorithm to measure the invariance of deep networks. In particular, it measures the robustness of deep networks to geometric transformations in a worst-case regime as they can be problematic for sensitive applications.

Pascal Frossard

Signal Processing Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Deep Neural Networks
PapersArticles
Source Code

SparseFool

Geometry-inspired sparse attack on deep networks

Deep Neural Networks have achieved extraordinary results on image classification tasks, but have been shown to be vulnerable to attacks with carefully crafted perturbations of the input data. Although most attacks usually change values of many image’s pixels, it has been shown that deep networks are also vulnerable to sparse alterations of the input. SparseFool implements an efficient algorithm to compute and control sparse alterations.

Pascal Frossard

Signal Processing Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Deep Neural Networks
PapersArticles
Source Code

Universal

Universal adversarial perturbations

Proposing a universal (image-agnostic) and very small perturbation vector that causes natural images to be misclassified with high probability in a deep neuronal network.

Pascal Frossard

Signal Processing Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Deep Neural Networks
PapersArticles
Source Code

Hold me tight!

Influence of discriminative features on deep network boundaries in ML

Important insights towards the explainability of neural networks reside in the characteristics of their decision boundaries. In this work, we borrow tools from the field of adversarial robustness, and propose a new perspective that relates dataset features to the distance of samples to the decision boundary. This enables us to carefully tweak the position of the training samples and measure the induced changes on the boundaries of CNNs trained on large-scale vision datasets. We use this framework to reveal some intriguing properties of CNNs. Specifically, we rigorously confirm that neural networks exhibit a high invariance to non-discriminative features, and show that the decision boundaries of a DNN can only exist as long as the classifier is trained with some features that hold them together. Finally, we show that the construction of the decision boundary is extremely sensitive to small perturbations of the training samples, and that changes in certain directions can lead to sudden invariances in the orthogonal ones. This is precisely the mechanism that adversarial training uses to achieve robustness.

Pascal Frossard

Signal Processing Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Deep Neural NetworksFeaturesInductive Bias
PapersArticles
Source Code

Neural Anisotropy Directions

Analyzing the role of the network architecture in shaping the inductive bias of deep classifiers.

In this work, we analyze the role of the network architecture in shaping the inductive bias of deep classifiers. To that end, we start by focusing on a very simple problem, i.e., classifying a class of linearly separable distributions, and show that, depending on the direction of the discriminative feature of the distribution, many state-of-the-art deep convolutional neural networks (CNNs) have a surprisingly hard time solving this simple task. We then define as neural anisotropy directions (NADs) the vectors that encapsulate the directional inductive bias of an architecture. These vectors, which are specific for each architecture and hence act as a signature, encode the preference of a network to separate the input data based on some particular features. We provide an efficient method to identify NADs for several CNN architectures and thus reveal their directional inductive biases. Furthermore, we show that, for the CIFAR-10 dataset, NADs characterize the features used by CNNs to discriminate between different classes.

Pascal Frossard

Signal Processing Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Deep Neural NetworksFeaturesInductive Bias
PapersArticles
Source Code

OrthoNet

Multilayer network data clustering

Network data appears in very diverse applications, like biological, social, or sensor networks. Clustering of network nodes into categories or communities has thus become a very common task in machine learning and data mining. Network data comes with some information about the network edges. In some cases, this network information can even be given with multiple views or multiple layers, each one representing a different type of relationship between the network nodes. Increasingly often, network nodes also carry a feature vector. We propose to extend the node clustering problem, that commonly considers only the network information, to a problem where both the network information and the node features are considered together for learning a clustering-friendly representation of the feature space.

Pascal Frossard

Signal Processing Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
FeaturesPyTorch
PapersArticles
Source Code

PRIME

Boost robustness of images to common corruptions

PRIME is a generic, plug-n-play data augmentation scheme that consists of simple families of max-entropy image transformations for conferring robustness to common corruptions. PRIME leads to significant improvements in corruption robustness on multiple benchmarks.

Pascal Frossard

Signal Processing Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
ImagesPyTorch
PapersArticles
Source Code

TransFool

Adversarial attack against neural machine translation models

Deep neural networks have been shown to be vulnerable to small perturbations of their inputs. In this paper, we investigate the vulnerability of Neural Machine Translation (NMT) models to these attacks and propose a new attack algorithm called TransFool. TransFool can severely degrade the translation quality for different translation tasks and NMT architectures. Moreover, we show that TransFool is transferable to unknown target models. Finally, based on automatic and human evaluations, TransFool leads to improvement in performance compared to the existing attacks. Thus, TransFool permits us to better characterize the vulnerability of NMT models and outlines the necessity to design strong defense mechanisms and more robust NMT systems for real-life applications.

Pascal Frossard

Signal Processing Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
AdversarialMachine LearningNatural Language
PapersArticles
Source Code

Adversarial Classification aTtack

Classification-Guided Adversarial Attack against NMT

Neural Machine Translation (NMT) models have been shown to be vulnerable to adversarial attacks, wherein carefully crafted perturbations of the input can mislead the target model. In this paper, we introduce ACT, a novel adversarial attack framework against NMT systems guided by a classifier. In our attack, the adversary aims to craft meaning-preserving adversarial examples whose translations in the target language by the NMT model belong to a different class (such as sentiment) than the original translations. Unlike previous attacks, our new approach has a more substantial effect on the translation by altering the overall meaning, which then leads to a different class determined by a classifier. Our attack is considerably more successful in altering the class of the output translation and has more effect on the translation. This new paradigm can reveal the vulnerabilities of NMT systems by focusing on the class of translation rather than the mere translation quality as studied traditionally.

Pascal Frossard

Signal Processing Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
AdversarialMachine LearningNatural Language
PapersArticles
Source Code

wehey

WeHe𝕐 - Localizing Traffic Differentiation

This project detects traffic differentiation using end-to-end throughput comparisons between an original network trace and a bit-inverted version that hides any packet payload patterns. It then attempts to localize where the differentiation occurs by analyzing measurements from additional paths that converge at the client's ISP. The localization relies on throughput comparison and loss trend correlation between these paths.

Katerina Argyraki

Network Architecture Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
AdversarialAnonymity
PapersArticles
Project Homepage
Source Code

tero project

Studying Internet latency from Gaming Footage

The system downloads gaming footage from the Twitch streaming platform. It extracts the displayed latency measurements using optical character recognition and knowledge of each game's user interface. It maps streamers to locations using natural language processing on their social media profiles. It analyzes the measurements to produce latency distributions per geographical location.

Katerina Argyraki

Network Architecture Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
AnonymityBenchmarkCloud
PapersArticles
Project Homepage
Source Code

gem5-X

gem5-based simulator with architectural extensions

gem5-X extends the gem5 full-system simulator with architectural eXtensions targeting heterogeneous many-core platforms, including custom memory hierarchies, power models, and hardware accelerator integration. It enables design-space exploration and optimization of HPC and embedded systems at the microarchitectural level using C++ and Python.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Simulator
PapersArticles
Project Homepage
Source Code

hdtorch

PyTorch-based hyperdimensional computing library

HDTorch is a PyTorch-based library designed for hyperdimensional computing, featuring custom CUDA extensions to accelerate hypervector operations. It is utilized for analyzing HDC benchmark datasets, focusing on accuracy, runtime, and memory usage. The workflow involves initializing basis vectors, encoding features, and aggregating vectors of the same class.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Machine LearningOptimizationTensorFlow
PapersArticles
Source Code

Bayesian Differential Privacy

Data distribution-aware differential privacy

Traditional differential privacy is independent of the data distribution. However, this is not well-matched with the modern machine learning con-text, where models are trained on specific data. As a result, achieving meaningful privacy guarantees in ML often excessively reduces accuracy. Bayesian differential privacy (BDP) takes into account the data distribution to provide more practical privacy guarantees.

Boi Faltings

Artificial Intelligence Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Anonymity
PapersArticles
Source Code

ADER

Continual adaption of recommendation systems without forgetting

Recommendation systems typically require continual adaptation to take into account new and obsolete items. A major challenge in this situation is catastrophic forgetting, where the trained model forgets patterns it has learned before. We propose a method to mitigate this effect.

Boi Faltings

Artificial Intelligence Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Decentralized
PapersArticles
Source Code

FeGAN

Scaling distributed Generative Adversarial Networks (GANs)

The FeGAN system enables training GANs in the Federated Learning setup. GANs are generative adversarial networks, a class of machine learning where two neural networks contest with each other. FeGAN is implemented on PyTorch and is using less bandwidth and is faster than using stat-of-the-art GANs implementations.

Rachid Guerraoui

Distributed Computing Lab

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
Federated LearningPyTorch
PapersArticles
Source Code

MVTIL

Multiversion Timestamp Locking (MVTL)

MVTL is a new family of concurrency control algorithms. Improving performance of distributed databases by allowing read accesses to lag behind the latest version of the data.

Rachid Guerraoui

Distributed Computing Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
ConcurrencyDatabaseLocking
PapersArticles
Source Code

libNVRAM

Log-free concurrent data structures

libNVRAM provides lock-free and lock-based concurrent data structures (linked lists, hash tables, skip lists) designed for non-volatile RAM (NVRAM). It eliminates the need for write-ahead logging by ensuring that data structures remain consistent across failures at the hardware level, achieving crash-safe persistence without the performance overhead of traditional journaling.

Rachid Guerraoui

Distributed Computing Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
ConcurrencyData StructuresFailure
PapersArticles
Project Homepage
Source Code

FloDB

Unlocking memory in persistent key-value stores

FloDB is a persistent key-value store that exploits large DRAM capacities on multi-core servers by maintaining a multi-level architecture: a small concurrent in-memory buffer feeds a larger DRAM-resident skip list, which is periodically flushed to an LSM-tree on persistent storage. This design decouples memory bandwidth from I/O bottlenecks, achieving high throughput on write-heavy workloads.

Rachid Guerraoui

Distributed Computing Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
DatabaseServer
PapersArticles
Project Homepage
Source Code

FLeet

Federated machine learning on Android

FLeet is a federated learning framework for Android that addresses the heterogeneity of mobile hardware by combining staleness-aware gradient aggregation with performance prediction. It selects and weights worker contributions based on predicted compute speed, reducing idle waiting and improving convergence speed and final model accuracy compared to synchronous federated learning baselines.

Rachid Guerraoui

Distributed Computing Lab

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
Federated LearningMobile App
PapersArticles
Source Code

MCTOP

A multi-core topology abstraction

MCTOP automatically constructs a hardware topology model for multi-core processors by measuring inter-core communication latencies and memory bandwidths at runtime. It exposes this model as a portable C API so that system software and concurrent data structure libraries can make topology-aware scheduling and memory placement decisions without requiring manual hardware configuration.

Rachid Guerraoui

Distributed Computing Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Server
PapersArticles
Project Homepage
Source Code

LOCKIN

Automatically chose the best locking algorithm

LOCKIN provides header-only implementations of multiple locking algorithms (MCS, CLH, TTAS, etc.) as drop-in replacements for pthreads mutexes. Its companion middleware GLS dynamically selects the optimal algorithm at runtime based on observed contention patterns, reducing energy consumption and improving throughput without requiring programmers to reason about low-level lock behaviour.

Rachid Guerraoui

Distributed Computing Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Concurrency
PapersArticles
Project Homepage
Source Code

AggregaThor

Framework over TensorFlow implementing robust stochastic gradient descent

AggregaThor is a distributed SGD framework layered over TensorFlow that integrates pluggable Byzantine-resilient gradient aggregation rules (e.g. Krum, Bulyan, coordinate-wise median) and supports asynchronous parameter-server topologies. It tolerates both Byzantine worker failures and unreliable network links, and its modular design allows aggregation rules and communication backends to be swapped independently.

Rachid Guerraoui

Distributed Computing Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Byzantine ResilienceDistributed LearningTensorFlow
PapersArticles
Source Code

ASCYLIB + OPTIK

State-of-the-art concurrent data-structure library

ASCYLIB is a comprehensive benchmark suite and library of over 40 concurrent search data structure implementations in C, covering linked lists, skip lists, hash tables, BSTs, queues, and priority queues in sequential, lock-based, and lock-free variants. OPTIK is a design pattern embedded within ASCYLIB that uses optimistic versioned locking to simplify the construction of correct and high-performance concurrent data structures.

Rachid Guerraoui

Distributed Computing Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
ConcurrencyData StructuresLocking
PapersArticles
Project Homepage
Source Code

CLHT

Very fast and scalable, resizable hash table

CLHT (Cache-Line Hash Table) is a highly optimised concurrent hash table in C whose buckets are aligned to single cache lines, eliminating false sharing. It provides both a lock-based variant with fine-grained per-bucket locking and a lock-free variant using hardware transactional memory, and supports online resizing while sustaining throughput above 50 million operations per second on commodity multi-core hardware.

Rachid Guerraoui

Distributed Computing Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
ConcurrencyData Structures
PapersArticles
Project Homepage
Source Code

Probabilistic Fault Tolerance of Neural Networks in the Continuous Limit

Regularizer for failing neurons

This work derives a closed-form probabilistic model of how neuron failures (permanent or transient) propagate through a deep neural network in the continuous-width limit. It uses this model to design a regulariser that, when added to standard training loss, encourages the network to learn weight distributions that are inherently robust to runtime neuron failures, without requiring fault injection during training.

Rachid Guerraoui

Distributed Computing Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Deep Neural Networks
PapersArticles
Source Code

uKharon

Membership service for microsecond applications

uKharon is a RDMA-based membership service that maintains a consistent view of active cluster members with sub-microsecond failure detection using RDMA one-sided operations and hardware timestamps. It separates the liveness oracle from the application data plane, allowing membership changes to be disseminated without entering the critical path of microsecond-scale request processing.

Rachid Guerraoui

Distributed Computing Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
ConcurrencyNetwork
PapersArticles
Source Code

uBFT

Microsecond-scale BFT using Disaggregated Memory

uBFT is a BFT state-machine replication protocol that achieves microsecond-scale latency by offloading agreement state to disaggregated RDMA memory accessible to all replicas. By replacing message-passing rounds with direct RDMA reads and writes, it reduces the critical-path latency of BFT consensus to under ten microseconds while maintaining standard safety and liveness guarantees under f Byzantine faults out of 3f+1 replicas.

Rachid Guerraoui

Distributed Computing Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Byzantine ResilienceConsensusNetworkProtocol
PapersArticles
Source Code

ffmalloc

Memory allocation preventing "use-after-free" exploits

ffmalloc is a "one time allocator" where any given virtual memory address is only returned to the calling application once. Since exploitation of use-after-free errors relies on being able to modify memory associated with "dangling pointers," by not reusing a virtual address region in subsequent allocation, an adversary cannot negatively impact the dangling memory region.

Sanidhya Kashyap

Robust Scalable Systems Software Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Protection
PapersArticles
Source Code

PACTree

Efficient Non-Volatile Memory (NVM) access

Non-Volatile Memory (NVM), which provides relatively fast and byte-addressable persistence, is now commercially available. However, we cannot equate a real NVM with a slow DRAM, as it is much more complicated than we expect. We propose guidelines for designing high-performance persistent index structures, allowing to increase performance and tail latency.

Sanidhya Kashyap

Robust Scalable Systems Software Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
ConcurrencyData Structures
PapersArticles
Source Code

NrOS

OS kernel with safer approach to synchronization

NrOS is an operating system primarily constructed as a simple, sequential kernel with no concurrency, making it easier to develop and reason about its correctness.

Sanidhya Kashyap

Robust Scalable Systems Software Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Concurrency
PapersArticles
Source Code

SynCord

Application-Informed Kernel Synchronization Primitives

SynCord is a framework to modify kernel locks without recompiling or rebooting the kernel. It abstracts key behaviors of kernel locks and exposes them as APIs for designing user-defined kernel locks. SynCord provides the mechanisms to customize kernel locks safely and correctly from userspace.

Sanidhya Kashyap

Robust Scalable Systems Software Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
ConcurrencyServer
PapersArticles
Project Homepage
Source Code

Odinfs

Scaling PM Performance with Opportunistic Delegation

Odinfs is a NUMA-aware PM file system based on an opportunistic delegation framework to achieve datapath scalability. It limits concurrent PM access to avoid performance meltdown due to on-DIMM cache trashing. It localizes PM access to avoid performance meltdown due to remote PM access. It automatically parallelizes large PM access to utilize cumulative PM bandwidth across all NUMA nodes. It maximizes concurrent access within the same file with scalable synchronization primitives.

Sanidhya Kashyap

Robust Scalable Systems Software Lab

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
ConcurrencyServer
PapersArticles
Project Homepage
Source Code

Hydra

Finding Semantic Bugs in File Systems with an Extensible Fuzzing Framework

Hydra is an extensible fuzzing framework for finding semantic bugs in file systems. It decouples the file system fuzzing pipeline into distinct components and provides abstract interfaces so researchers can plug in custom bug oracles, mutators, and executors. Hydra has been used to discover numerous real bugs in Linux file systems including ext4, btrfs, F2FS, and XFS.

Sanidhya Kashyap

Robust Scalable Systems Software Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
ConcurrencyFuzzingServer
PapersArticles
Source Code

TCLocks

Scalable transparent delegation locking protocols without application modification

TCLocks is a family of locking protocols that use transparent delegation to improve scalability of synchronization in multi-core systems. The waiter encodes its critical section on its stack and notifies the lock holder (combiner), which executes the critical section on the waiter's behalf via a light-weight context switch. This avoids transferring shared data between cores. TCLocks require no modifications to application logic and provide up to 5.2x better performance than recent locking algorithms.

Sanidhya Kashyap

Robust Scalable Systems Software Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
ConcurrencyDevelopmentLockingServer
PapersArticles
Source Code

Trio Architecture

Trio enables customizable user-level file systems with metadata integrity

Trio is a userspace NVM file system architecture that decouples file system design, access control, and metadata integrity enforcement. It defines a core state data structure shared between LibFSes and a trusted entity. LibFSes can customize with private auxiliary state while directly accessing NVM. The trusted entity verifies core state integrity when write access transfers between LibFSes. This allows customization while ensuring metadata integrity. Trio includes ArckFS, a POSIX-like FS, and customized FSes that outperform existing NVM file systems.

Sanidhya Kashyap

Robust Scalable Systems Software Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
CloudDevelopmentLow-LevelServer
PapersArticles
Source Code

ADEFAN

Fast and energy-efficient video encoder and decoder

ADEFAN (Adaptive Depth Efficient Filtering by Adaptive Normaliztion) is a state-of-the-art technology for encoding and decoding videos that presents many advantages over other available methods. It features extremely energy-efficient encoder, extremely fast encoding process, no specialized hardware for encoding and added security with encoding key. It is this ideal for video-transmitting IoT devices, continuously transmitting surveillance cameras, emergency deployment in areas without much infrastructure (remote, natural disaster relief, etc.).

Sabine Süsstrunk

Image and Visual Representation Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Computer VisionImages
PapersArticles
Project Homepage
Source Code

FG-NIC

Noise-robust image classification

Image classification has significantly improved using deep learning. This is mainly due to convolutional neural networks (CNNs) that are capable of learning rich feature extractors from large datasets. However, most deep learning classification methods are trained on clean images and are not robust when handling noisy ones, even if a restoration preprocessing step is applied. We propose a method that can be applied on a pretrained classifier. We improve the noisy-image classification (NIC) results by significantly large margins, especially at high noise levels, and come close to the fully retrained approaches.

Sabine Süsstrunk

Image and Visual Representation Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Deep Neural NetworksImage ClassificationImages
PapersArticles
Source Code

BUIFD

Deep learning image denoiser

Blind and universal image denoising consists of using a unique model that denoises images with any level of noise. It is especially practical as noise levels do not need to be known when the model is developed or at test time. We propose a blind and universal deep learning image denoiser based on an optimal denoising solution, which we call fusion denoising. Our approach improves real-world grayscale additive image denoising. It also improves state-of-the-art color image denoising performance on every single noise level by an average of 0.1dB.

Sabine Süsstrunk

Image and Visual Representation Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Deep Neural NetworksImages
PapersArticles
Source Code

Zero-Learning Fast Medical Image Fusion

High-quality image fusion

Medical image fusion plays a central role by integrating information from multiple sources into a single, more understandable output. We propose a real-time image fusion method using pre-trained neural networks to generate a single image containing features from multi-modal sources. Our method can be applied to any number of input sources.

Sabine Süsstrunk

Image and Visual Representation Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Deep Neural NetworksImagesMedical
PapersArticles
Source Code

LELSD

Interactive GAN-based image editing

When generating an image using a GAN, some errors might appear which are obvious to the human observer, but not to the computer. By introducing this user feedback one can refine the generated image to look even more real.

Sabine Süsstrunk

Image and Visual Representation Lab

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
Deep Neural NetworksImages
PapersArticles
Source Code

DSR

Drone Image Super-Resolution

Despite achieving remarkable progress in recent years, single-image super-resolution methods are developed with artificial limitations. We propose a novel drone image dataset, with scenes captured at low and high resolutions, and across a span of altitudes. Our results show that off-the-shelf state-of-the-art networks perform worse on this different domain.

Sabine Süsstrunk

Image and Visual Representation Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Deep Neural NetworksImages
PapersArticles
Source Code

TempSAL

Uncovering Temporal Information for Deep Saliency Prediction

Deep saliency prediction algorithms complement the object recognition features, they typically rely on additional information, such as scene context, semantic relationships, gaze direction, and object dissimilarity. However, none of these models consider the temporal nature of gaze shifts during image observation. We introduce a novel saliency prediction model that learns to output saliency maps in sequential time intervals by exploiting human temporal attention patterns.

Sabine Süsstrunk

Image and Visual Representation Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Deep Neural NetworksImages
PapersArticles
Project Homepage
Source Code

DyNCA

Real-Time Dynamic Texture Synthesis Using Neural Cellular Automata

Current Dynamic Texture Synthesis (DyTS) models in the literature can synthesize realistic videos. However, these methods require a slow iterative optimization process to synthesize a single fixed-size short video, and they do not offer any post-training control over the synthesis process. We propose Dynamic Neural Cellular Automata (DyNCA), a framework for real-time and controllable dynamic texture synthesis.

Sabine Süsstrunk

Image and Visual Representation Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Deep Neural NetworksImages
PapersArticles
Project Homepage
Source Code

Gotee

GoLang extension to allow a programmer to execute a goroutine within an enclave

Trusted Execution Environments (TEEs), such as Intel SGX enclaves, use hardware to ensure the confidentiality and integrity of operations on sensitive data. While the technology is available on many processors, the complexity of its programming model and its performance overhead have limited adoption. TEEs provide a new and valuable hardware functionality that has no obvious analogue in programming languages, which means that developers must manually partition their application into trusted and untrusted components. Gotee extends the Go language to allow a programmer to execute a goroutine within an enclave, to use low-overhead channels to communicate between the trusted and untrusted environments, and to rely on a compiler to automatically extract the secure code and data.

Edouard Bugnion

Data Center Systems Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
DevelopmentProtection
PapersArticles
Source Code

Raft consensus protocol

Implementation of the Raft consensus protocol

This is a C implementation of the Raft distributed consensus protocol that deliberately makes no assumptions about the underlying network, so it can operate over unreliable transports such as UDP. The library exposes a state-machine interface for log replication and leader election, allowing integrators to plug in their own network and storage layers.

Edouard Bugnion

Data Center Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
ConsensusDecentralizedProtocol
PapersArticles
Source Code

HovercRaft

Improved scalable and fault-tolerant distributed architecture

Cloud platform services must simultaneously be scalable, meet low tail latency service-level objectives, and be resilient to a combination of software, hardware, and network failures. Replication plays a fundamental role in meeting both the scalability and the fault-tolerance requirement, but is subject to opposing requirements: (1) scalability is typicallyachieved by relaxing consistency; (2) fault-tolerance is typically achieved through the consistent replication of state machines. Adding nodes to a system can therefore either increase performance at the expense of consistency, or increase resiliency at the expense of performance. We propose HovercRaft, a new approach by which adding nodes increases both the resilience and the performance of general-purpose state-machine replication.

Edouard Bugnion

Data Center Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Byzantine ResilienceConsensusProtocol
PapersArticles
Source Code

CRAB

L4 load balancer for cloud provider

Connection Redirect LoAd Balancer is a new scheme for L4 load balancing targeting specifically internal cloud workloads.

Edouard Bugnion

Data Center Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
CloudNetwork
PapersArticles
Source Code

Rust-Stainless

Rust frontend for Stainless

Rust-Stainless is a frontend that extracts a Stainless-compatible intermediate representation from Rust source code via the Rust compiler internals. It enables deductive verification of safe Rust programs by translating annotated functions into Stainless verification conditions and dispatching them to the underlying SMT-based solver.

Viktor Kunčak

Lab for Automated Reasoning and Analysis

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
Code AnalysisStatic Analysis
PapersArticles
Source Code
Contact

SecVM

Privacy-preserving classification

SecVM trains a Support Vector Machine (SVM) classifier in a federated setting where the central server is considered untrusted. User data is kept local and only encrypted gradient updates are sent to the server, using secret-sharing and homomorphic-encryption-inspired techniques to prevent the server from learning individual data points or their labels.

Robert West

Data Science Lab

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
DecentralizedDistributed Learning
PapersArticles
Source Code

Invariant Language Modeling

Invariant natural language modeling

This framework adapts Invariant Risk Minimization (IRM) to natural language tasks by defining multiple training environments from available data and penalizing features whose predictive power varies across environments. It integrates with standard pretrained Transformer encoders and demonstrates improved robustness on out-of-domain NLI and sentiment benchmarks.

Robert West

Data Science Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Natural Language
PapersArticles
Source Code

Eigenthemes

Improved entity linking

Eigenthemes performs unsupervised entity linking by exploiting the observation that entities mentioned in the same document tend to cluster in a low-dimensional subspace of a pretrained entity embedding space. It extracts this subspace via SVD and ranks entity candidates by their projection onto the dominant singular vectors, without requiring any labeled training data.

Robert West

Data Science Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Natural Language
PapersArticles
Source Code

Quotebank

Corpus of quotations from a decade of news

Quotebank is a large-scale dataset of 235 million quotations extracted from 162 million English news articles published between 2008 and 2020. Speaker attribution is performed using a Wikidata-linked entity linking pipeline and a probabilistic model trained on Wikipedia-derived supervision, yielding a quotation-to-speaker corpus suitable for NLP, social science, and computational journalism research.

Robert West

Data Science Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Natural Language
PapersArticles
Source Code

DiPPS

Differentially Private Propensity Scores for Bias Correction

DiPPS estimates propensity scores under differential privacy to reweight a biased sample toward the true population distribution. It uses a private logistic regression model to compute importance weights, then applies them for unbiased downstream statistical estimation. The approach handles both voluntary participation bias and distribution shift from proxy datasets, with formal differential privacy guarantees.

Robert West

Data Science Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Survey
PapersArticles
Source Code

GenIE

Autoregressive information extraction system

GenIE fine-tunes a seq2seq Transformer (BART) for closed information extraction by autoregressively generating entity-linked (subject, relation, object) triplets constrained to a target knowledge base. A constrained decoding procedure enforces that all generated entities and relations are valid Wikidata identifiers, yielding fully grounded and structured output from raw text.

Robert West

Data Science Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Natural Language
PapersArticles
Source Code

SynthIE

Exploiting Asymmetry for Synthetic Training Data Generation

SynthIE leverages the asymmetry between text-to-triples and triples-to-text: it uses a large language model to generate fluent text from Wikidata (subject, relation, object) triples, creating a large synthetic corpus for training information-extraction models. The resulting dataset bootstraps high-quality IE models without requiring expensive manual annotation, and outperforms models trained on existing human-labeled corpora.

Robert West

Data Science Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Natural Language
PapersArticles
Source Code

aiFlows

Modular AI collaboration framework

aiFlows is a framework centered around Flows and messages. Flows are independent, self-contained computational building blocks that can complete semantically meaningful units of work. Flows communicate via a standardized message-based interface, enabling modularity, reusability, remote peer-to-peer collaboration, and concurrency.

Robert West

Data Science Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
CloudMachine LearningProtocol
PapersArticles
Project Homepage
Source Code

Transformers CFG

Grammar-constrained text generation with Transformers models

Transformers_cfg is an extension library for the Hugging Face Transformers library that enables grammar-constrained text generation. It provides tools and functionalities to work with context-free grammars (CFGs) for natural language processing tasks involving CFGs. The library supports various Transformer models, including LLaMa, GPT, Bloom, Mistral, and Falcon, and offers features like multilingual grammar support and integration with Text-Generation-WebUI.

Robert West

Data Science Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Machine LearningNatural Language
PapersArticles
Source Code

Entity Insertion in Wikipedia

Multilingual entity insertion in Wikipedia articles

Proposes a framework for inserting entities into Wikipedia articles across multiple languages. It processes Wikipedia dumps to extract data and train models for entity insertion. The key components are: 1) Data processing pipeline to extract relevant data from Wikipedia dumps. 2) Modeling code for training entity insertion models using a ranking loss or pointwise loss. 3) Benchmarking code to evaluate models against baselines like BM25, EntQA, and GPT language models.

Robert West

Data Science Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Machine LearningNatural Language
PapersArticles
Source Code

LAMEN

Evaluating language models through negotiation tasks

The project proposes using structured negotiations as a dynamic benchmark for evaluating language model (LM) agents. The negotiation framework consists of a game setting, issues to negotiate, and optional preference weights, allowing for the design of complex games by increasing the number of issues, mixing issue types, and adding non-uniform preferences. The benchmark setup jointly evaluates performance metrics (utility and completion rate) and alignment metrics (faithfulness and instruction-following) in self-play and cross-play settings.

Robert West

Data Science Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Machine LearningNatural LanguageOptimization
PapersArticles
Project Homepage
Source Code

LLM Grounding Analysis

LLM grounding vs. factual recall

The study investigates LLMs using Fakepedia, a dataset presenting contradictions between known facts and new information. Through Masked Grouped Causal Tracing (MGCT), the research deciphers LLMs' grounding mechanisms by contrasting neural activation patterns. Findings help understand the co-functioning of grounding with recall capabilities within LLMs.

Robert West

Data Science Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Machine LearningNatural Language
PapersArticles
Source Code

The Open Food Repo

Database of barcoded food products

Open Food Repo is a crowdsourced, openly licensed database of barcoded food products available in Switzerland, exposed via a documented REST API. It stores product names, ingredients, nutritional values, and images, and serves as the product data backbone for the MyFoodRepo dietary tracking platform.

Marcel Salathé

Salathe Lab

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
Database
PapersArticles
Project Homepage
Source Code
Contact

CrowdBreaks

Track health trends using public social media data and crowdsourcing

CrowdBreaks is an open platform that combines streaming ingestion of public social media data with crowdsourced labeling and automated machine learning pipelines to track population-level health trends over time. It provides real-time dashboards of sentiment and topic trends and is designed to be extended to new health surveillance use cases.

Marcel Salathé

Salathe Lab

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
CrowdSourceMedicalSurvey
PapersArticles
Project Homepage
Source Code
Contact

txcl

Reproducible Text Classification

txcl is a Python library for reproducible text classification that wraps standard NLP workflows including data loading, preprocessing, fine-tuning of transformer-based models, and evaluation. It provides a configuration- driven interface to support consistent experimental setups and was developed in the context of social media health surveillance research.

Marcel Salathé

Salathe Lab

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
CrowdSourceNatural Language
PapersArticles
Source Code

CloudSuite

Benchmark suite for cloud services

CloudSuite is a benchmark suite for cloud services. It consists of a number of applications that have been selected based on their popularity in today's datacenters. The benchmarks are based on real-world software stacks and represent real-world setups.

Babak Falsafi

Parallel Systems Architecture Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
BenchmarkCloud
PapersArticles
Project Homepage
Source Code

ColTraIn HBFP Training Emulator

Co-located deep neural network training and inference

HBFP is a hybrid Block Floating-Point (BFP) - Floating-Point (FP) number representation for DNN training introduced by ColTraIn: Co-located DNN Training and Inference team of PARSA and MLO at EPFL. HBFP offers the best of both worlds: the high accuracy of floating-point at the superior hardware density of fixed-point by performing all dot products in BFP and other operations in FP32. For a wide variety of models, HBFP matches floating-point’s accuracy while enabling hardware implementations that deliver up to 8.5x higher throughput. This repository is for ongoing research on training DNNs with HBFP.

Babak Falsafi

Parallel Systems Architecture Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Deep Neural Networks
PapersArticles
Project Homepage
Source Code
Contact

Unlynx

Privacy-preserving data sharing

Unlynx is a decentralized system for privacy-preserving data sharing built on top of the Cothority framework. It provides composable cryptographic protocols — including collective aggregation, key switching, and oblivious shuffling — based on homomorphic encryption and secure multi-party computation, with verifiable correctness guarantees.

Jean-Pierre Hubaux

Laboratory for Data Security

Maturity
Support
C4DT
Inactive
Lab
Unknown
Homomorphic EncryptionMedicalProtocolSecure Multi-Party Computation
PapersArticles
Source Code

PriCell

Privacy preserving neural network learning for diseased cell classification

PriCell is a Golang library that ensures CellCnn analysis in privacy-preserving, distributed, N-party setting.

Jean-Pierre Hubaux

Laboratory for Data Security

Maturity
Support
C4DT
Inactive
Lab
Unknown
Homomorphic EncryptionMedical
PapersArticles
Source Code

CRISP

Offload computation to services while keeping data private

CRISP relies on quantum-resistant lattice-based approximate homomorphic encryption primitives that support flexible polynomial computations on encrypted data without degrading utility. To ensure data integrity, it employs lattice-based commitments and zero-knowledge proofs based on the multi-party-computation-in-the-head paradigm, which enables users to simultaneously convince service providers about the correctness of the encrypted data as well as the authenticity of the underlying plaintext data, using the deployed certification mechanism.

Jean-Pierre Hubaux

Laboratory for Data Security

Maturity
Support
C4DT
Inactive
Lab
Unknown
Homomorphic EncryptionZero-Knowledge Proofs
PapersArticles
Source Code

peacok

Persona Commonsense Knowledge for Consistent and Engaging Narratives

PeaCoK is a large-scale persona commonsense knowledge graph, containing approximately 100K human-validated persona facts. It schematizes five dimensions of persona knowledge identified in previous studies of human interactive behaviours, and distils facts in this schema from both existing commonsense knowledge graphs and large-scale pretrained language models. The analysis indicates that PeaCoK contains rich and precise world persona inferences that help downstream systems generate more consistent and engaging narratives.

Antoine Bosselut

Natural Language Processing Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Machine LearningNatural Language
PapersArticles
Source Code

distilled counterfactual data

Automated generation of high-quality counterfactual data.

DISCO (DIStilled COunterfactual Data) is a method for automatically generating high-quality counterfactual data at scale. It uses a large general language model to generate phrasal perturbations, which are then filtered by a task-specific teacher model to distill high-quality counterfactual data. The method has been applied to natural language inference tasks, demonstrating improved robustness and generalization across distributions.

Antoine Bosselut

Natural Language Processing Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
AdversarialMachine LearningNatural Language
PapersArticles
Source Code

meditron

Medical language model pretraining

MEDITRON-70B is a large language model (LLM) aimed at democratizing access to medical knowledge. It is an open-source LLM with 7B and 70B parameters, adapted specifically for the medical domain. The model builds on Llama-2 and extends pretraining on a curated medical corpus, including selected PubMed articles and internationally-recognized medical guidelines.

Antoine Bosselut

Natural Language Processing Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Machine LearningMedicalNatural Language
PapersArticles
Source Code

crow

Benchmarking Commonsense Reasoning in Real-World Tasks

CRoW is a manually-curated, multi-task benchmark that evaluates the ability of models to apply commonsense reasoning in the context of six real-world NLP tasks. It is constructed using a multi-stage data collection pipeline that rewrites examples from existing datasets using commonsense-violating perturbations. The study reveals a significant performance gap when NLP systems are evaluated on CRoW compared to humans, indicating that commonsense reasoning is far from being solved in real-world task settings.

Antoine Bosselut

Natural Language Processing Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
BenchmarkMachine LearningNatural Language
PapersArticles
Project Homepage
Source Code

JumboTrace

Java program tracer to record execution state

JumboTrace is an experimental tracer for Java programs that records the evolution of program state during execution to build a trace (sequence of events and executed lines). It works by injecting logging calls into the program using a compiler plugin for javac. The logged events are written to a binary file using serialization, and a minimal frontend can display the traces in the console.

Clément Pit-Claudel

Systems and Formalisms Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Code AnalysisDevelopment
PapersArticles
Source Code

DynaBurst

Group DRAM bursts

DynaBurst is a highly flexible, FPGA-optimized, memory system for bandwidth-bound accelerators that perform frequent irregular accesses to DRAM memory. DynaBurst is an extension of the lab's miss-optimized nonblocking cache which groups incoming requests into bursts of memory requests.

Paolo Ienne

Processor Architecture Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Low-Level
PapersArticles
Source Code

iprp

Parallel redundancy protocol for IP networks

iPRP allows parallel transmission of data over a network for hard real time applications with simply installing an app on both the server and the client.

Jean-Yves Le Boudec

Computer Communications and Applications Laboratory 2

Maturity
Support
C4DT
Inactive
Lab
Unknown
Network
PapersArticles
Source Code

commelec-api

API for the Commelec real-time microgrid framework

This project defines a message format, based on Cap'n Proto, for encoding certain mathematical objects (i.e., real-valued symbolic expressions, sets and set-valued functions) as sequences of bytes.

Jean-Yves Le Boudec

Computer Communications and Applications Laboratory 2

Maturity
Support
C4DT
Inactive
Lab
Unknown
Low-LevelNetwork
PapersArticles
Source Code
Contact

T-RECS

Testbed for multi-agent real-time control of electric grids

T-RECS is a software testbed for multi-agent real-time control of electric grids. T-RECS uses software containers to run each agent in a virtual environment. Agents are networked via Mininet, so real packets are exchanged using TCP/IP.

Jean-Yves Le Boudec

Computer Communications and Applications Laboratory 2

Maturity
Support
C4DT
Inactive
Lab
Unknown
Network
PapersArticles
Source Code

Synthetic data privacy evaluation

Privacy evaluation framework for synthetic data publishing

The framework implemented in this library measures the privacy gain of publishing a synthetic dataset in place of the raw data with respect to a specific privacy concern. Each concern is modelled as a privacy adversary that targets an individual record and aims to infer a secret about this record. The library includes implementations of two new privacy attacks on the output of a generative model. To evaluate privacy gain, the framework is instantiated under the chosen threat model and outputs an estimate about how much publishing the synthetic data instead of the raw data reduces the privacy loss of a chosen target record under this threat model.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Protocol
PapersArticles
Source Code
Contact

ClaimChain

Prove your identity with regards to others

A decentralized public key distribution system with an emphasis on privacy preservation.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Decentralized
PapersArticles
Project Homepage
Source Code

Lightnion

Anonymous communication in browsers via the Tor network

A Javascript library for browsers for anonymous communication via the Tor network. The library connects to the Tor network to enable websites to send and retrieve data anonymously (e.g. send sensitive information without identifying the user) without requiring users to install special software such as the Tor browser.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
NetworkTor
PapersArticles
Source Code

Trickster

Sophisticated attacks on Machine Learning models

Trickster allows to attack machine learning models where the attack needs to be more sophisticated than simply adding some noise to an image. One example is to evade an abuse detection model, for example, social media bot detection, or a malware detector.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Attack
PapersArticles
Source Code

Hippiepug

Sublinear-traversal blockchains and efficient key-value Merkle trees

This library provides implementations of two cryptographic data structures: fast traversal blockchain structure, and a verifiable dictionary, implemented as a key-value Merkle tree.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Decentralized
PapersArticles
Source Code

MCSAuditing

Evaluation code for mobile crowdsourcing privacy

This tool implements the evaluation framework described in the paper

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Decentralized
PapersArticles
Source Code

Zero-Knowledge Swiss Knife

Library for prototyping composable zero-knowledge proofs in the discrete-log setting

A Python library for prototyping and composing sigma-protocol-based zero-knowledge proofs in the discrete-logarithm setting. It supports conjunctions, disjunctions, and custom proof statements, enabling rapid construction of complex proofs for research and prototyping purposes.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
Zero-Knowledge Proofs
PapersArticles
Source Code

Petrelic

Wrapper around RELIC pairings

petrelic is a Python wrapper around RELIC. It provides a simple python interface to the BLS-381 pairing and RELIC's big number class.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Low-Level
PapersArticles
Source Code

SSCred

Library to facilitate anonymous authentication

SSCred provides the following primitives: 1) Anonymous credential light (ACL) 2) Abe's Blind signature 3) Blinded Pedersen commitment

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Anonymity
PapersArticles
Source Code

mia

Library for running membership inference attacks (MIA) against machine learning models

These are attacks against privacy of the training data. In MIA, an attacker tries to guess whether a given example was used during training of a target model or not, only by querying the model. See more in the paper by Shokri et al. Currently, you can use the library to evaluate the robustness of your Keras or PyTorch models to MIA.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Attack
PapersArticles
Source Code

VOTEAGAIN

Scalable coercion-resistant voting system

The strongest threat model for voting systems considerscoercion resistance: protection against coercers that force voters to modify their votes, or to abstain. Existing remote voting systems either do not provide this property; require an ex-pensive tallying phase; or burden users with the need to store cryptographic key material and with the responsibility to deceive their coercers. VoteAgain is a scalable voting scheme that relies on the revoting paradigm to provide coercion resistance.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Anonymity
PapersArticles
Source Code

DatashareNetwork

Decentralized privacy-preserving search engine for investigative journalists

Investigative journalists collect large numbers of digital documents during their investigations. These documents can greatly benefit other journalists’ work. However, many of these documents contain sensitive information. Hence, possessing such documents can endanger reporters, their stories, and their sources. Consequently, many documents are used only for single, local, investigations. DatashareNetwork is a decentralized and privacy-preserving search system that enables journalists world-wide to find documents via a dedicated network of peers.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Decentralized
PapersArticles
Source Code

Tandem

Centralized privacy-preserving secure key storage

Users’ devices, e.g., smartphones or laptops, are typically incapable of securely storing and processing cryptographic keys. Tandem is a novel set of protocols for securing cryptographic keys with support from a central server.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
AnonymityEncryptionProtocol
PapersArticles
Source Code

WebGraph

Capturing Advertising and Tracking Information Flows for Robust Blocking

ML-based ad and tracker blocker that detects ads and trackers based on their action rather than their content.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
AnonymityProtection
PapersArticles
Source Code

DisparateVulnerability

Disparate Vulnerability to Membership Inference Attacks

A membership inference attack (MIA) against a machine-learning model enables an attacker to determine whether a given data record was part of the model's training data or not. DisparateVulnerability helps in measuring how much a given model can be inverted.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
AnonymityProtection
PapersArticles
Source Code

offchainer

Ethereum smart contracts with secret parameters

By design, smart contracts’ data and computations are public to all participants. Offchainer helps to create smart contracts with parameters that need to stay secret. We propose a way to keep some of the parameters off-chain, while guaranteeing correctness of the computation, using a combination of a commitment scheme and a zero-knowledge proof system.

Serge Vaudenay

Security and Cryptography Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
DecentralizedZero-Knowledge Proofs
PapersArticles
Source Code

ophcrack

Windows password cracker

Ophcrack is a Windows password cracker based on rainbow tables. It is a very efficient implementation of rainbow tables done by the inventors of the method. It comes with a Graphical User Interface and runs on multiple platforms

Serge Vaudenay

Security and Cryptography Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Attack
PapersArticles
Project Homepage
Source Code

non/transferable anonymous tokens

Anonymous tokens with non-transferability and accountability

We developed techniques for building up anonymous tokens. NTAT is non-transferable in the sense that a client who received an anonymous voucher from an issuer cannot sell it to another client without leaving a valuable secret with it.

Serge Vaudenay

Security and Cryptography Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Access ControlAnonymity
PapersArticles
Source Code

STORM+

Faster PyTorch optimizer

STORM+ is a fully adaptive stochastic gradient descent optimizer with momentum for nonconvex optimization, extending the STORM algorithm. It achieves optimal convergence rates without requiring knowledge of problem-specific parameters such as the Lipschitz constant or noise level, at the cost of two gradient evaluations per iteration.

Volkan Cevher

Laboratory for Information and Inference Systems

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
PyTorch
PapersArticles
Source Code

PowerGossip

Practical low-rank communication compression in decentralized deep learning

Inspired by the PowerSGD algorithm for centralized deep learning, this algorithm uses power iteration steps to maximize the information transferred per bit. We prove that our method requires no additional hyperparameters, converges faster than prior methods, and is asymptotically independent of both the network and the compression.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Deep Neural Networks
PapersArticles
Source Code

chocoSGD

Decentralized communication-efficient machine learning and deep learning

Communication-efficient decentralized ML training (both deep learning, compatible with PyTorch, and traditional convex machine learning models).

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
DecentralizedDistributed LearningPyTorch
PapersArticles
Source Code

cola

Decentralized linear machine learning

Decentralized machine learning is a promising emerging paradigm in view of global challenges of data ownership and privacy. We consider learning of linear classification and regression models, in the setting where the training data is decentralized over many user devices, and the learning algorithm must run on-device, on an arbitrary communication network, without a central coordinator. We propose COLA, a new decentralized training algorithm with strong theoretical guarantees and superior practical performance. Our framework overcomes many limitations of existing methods, and achieves communication efficiency, scalability, elasticity as well as resilience to changes in data and participating devices.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
DecentralizedDistributed Learning
PapersArticles
Source Code
Contact

sent2vec

Numerical representations for short texts

Library that delivers numerical representations (features) for words, short texts or sentences, which can be used as input to any machine learning task later on.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
Natural Language
PapersArticles
Source Code
Contact

PowerSGD

Practical low-rank gradient compression for distributed optimization

New low-rank gradient compressor based on power iteration that can i) compress gradients rapidly, ii) efficiently aggregate the compressed gradients using all-reduce, and iii) achieve test performance on par with SGD. The proposed algorithm is the only method evaluated that achieves consistent wall-clock speedups when benchmarked against regular SGD with an optimized communication backend. We demonstrate reduced training times for convolutional networks as well as LSTMs on common datasets.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
DecentralizedDistributed Learning
PapersArticles
Source Code

RelaySGD

Improved information propagation in decentralized learning

Because the workers only communicate with few neighbors without central coordination, these updates propagate progressively over the network. This paradigm enables distributed training on networks without all-to-all connectivity, helping to protect data privacy as well as to reduce the communication cost of distributed training in data centers. A key challenge, primarily in decentralized deep learning, remains the handling of differences between the workers' local data distributions. To tackle this challenge, we introduce the RelaySum mechanism for information propagation in decentralized learning. RelaySum uses spanning trees to distribute information exactly uniformly across all workers with finite delays depending on the distance between nodes. In contrast, the typical gossip averaging mechanism only distributes data uniformly asymptotically while using the same communication volume per step as RelaySum.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
DecentralizedDistributed LearningPyTorch
PapersArticles
Source Code

Byzantine Robust Optimizer

Improved federated learning with Byzantine robustness

Byzantine robustness has received significant attention recently given its importance for distributed and federated learning. In spite of this, there are severe flaws in existing algorithms even when the data across the participants is identically distributed. To address these issues, we present two surprisingly simple strategies: a new robust iterative clipping procedure, and incorporating worker momentum to overcome time-coupled attacks. This is the first provably robust method for the standard stochastic optimization setting.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
DecentralizedDistributed LearningTensorFlow
PapersArticles
Source Code
Contact

Quasi-Global Momentum

Improved decentralized training with data heterogeneity

Decentralized training of deep learning models is a key element for enabling data privacy and on-device learning over networks. In realistic learning scenarios, the presence of heterogeneity across different clients' local datasets poses an optimization challenge and may severely deteriorate the generalization performance. We propose a novel momentum-based method to mitigate this decentralized training difficulty.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
DecentralizedDeep Neural NetworksDistributed Learning
PapersArticles
Source Code

HyperAggregate

Sublinear aggregation algorithm

Novel decentralized aggregation protocol that can be parameterized so that the overall computation overhead scales logarithmically with the number of nodes.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
DecentralizedDistributed Learning
PapersArticles
Source Code

PhantoMedicus

Medical Survey Generator

Phantomedicus is a framework for simulating patients and consultations. It can extract disease probabilities from given data.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Predictor
PapersArticles
Source Code

PAXlib

Augmentent PyTorch with some JAX

PAXlib augments PyTorch with JAX-inspired abstractions including functional state management and pytree-compatible module structures. It enables researchers to write more modular and composable deep learning code within a PyTorch workflow, without requiring a full migration to JAX.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
PyTorch
PapersArticles
Source Code

Byzantine-Robust NonIID optimizer

Byzantine-Robust Learning on Heterogeneous Datasets via Bucketing

An implementation of a bucketing-based Byzantine-robust aggregation strategy for federated learning with heterogeneous (non-IID) data distributions. The bucketing technique reduces the effective variance introduced by data heterogeneity, enabling standard robust aggregation rules to function correctly in the non-IID setting. Evaluated against state-of-the-art attacks and baselines in PyTorch.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
PyTorch
PapersArticles
Source Code

Distributed Homomorphic Anomaly Detection

Proof-of-concept ML algorithms to do anomaly detection compatible with distributed, encrypted algorithms.

The goal of this project is prove that state-of-the-art anomaly detection algorithms that already achieve over 99% accuracy can be run in a distributed and encrypted fashion. In other words, prove that multiple parties can collectively train a good anomaly detecting model without revealing their data to each other. This has been achieved through transfer learning into a Multi-layer Perceptron (MLP) model which achieved comparable results (95% compared to 99% False-Positive Rate accuracy).

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Distributed LearningHomomorphic EncryptionPyTorch
PapersArticles
Source Code

Meditron

Open-source medical language model for clinical decision support

MEDITRON is a pair of open-source large language models (LLMs) with 7 and 70 billion parameters, tailored to the medical domain. It was trained on carefully curated medical data sources, including peer-reviewed literature and clinical practice guidelines. MEDITRON outperforms other open-source models and closed models like GPT-3.5 on medical benchmarks, coming within 5-10% of GPT-4 and Med-PaLM-2.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Machine LearningNatural Language
PapersArticles
Project Homepage
Source Code

Megatron-LLM

Large language model training library

Megatron-LLM enables pre-training and fine-tuning of large language models (LLMs) at scale. It supports architectures like Llama, Llama 2, Code Llama, Falcon, and Mistral. The library allows training of large models (up to 70B parameters) on commodity hardware using tensor, pipeline, and data parallelism. It provides features like grouped-query attention, rotary position embeddings, BF16/FP16 training, and integration with Hugging Face and WandB.

Martin Jaggi

Machine Learning and Optimization Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Machine LearningNatural Language
PapersArticles
Source Code

BOLT

Performance contracts for software network functions

We propose the notion of performance contracts for NFs. A performance contract enables users to predict and scrutinize NF performance at fine granularities without having to run them. It provider users with an abstraction that enables them to easily parameterize arbitrary input workloads, whether typical, exceptional or adversarial. Given only this abstract description of the workload and not a concrete instance, the contract predicts the performance of the NF without actually running it. The performance predictions are in terms of human-readable formulae, expressed as a function of workload and environment variables, that we together call Performance Critical Variables (PCVs). informed development decisions.

George Candea

Dependable Systems Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Network
PapersArticles
Project Homepage
Source Code
Contact

PIX

Automatic extraction of performance interfaces for network functions

PIX automatically extracts performance interfaces from network function (NF) implementations. A performance interface characterises an NF's throughput and latency as a function of workload and environment variables, giving a compact and accurate summary of its performance profile. Evaluated on 12 network functions, PIX produces interfaces that are accurate yet far simpler than the source code, and extracts them in minutes. Practical applications include detecting performance regressions, diagnosing performance bugs, and measuring the latency impact of NIC offloads.

George Candea

Dependable Systems Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Code AnalysisLow-LevelNetwork
PapersArticles
Source Code

eBPF-SE

eBPF symbolic execution tool

eBPF-SE is a tool that symbolically executes eBPF programs written for the Linux kernel. It is based on the KLEE symbolic execution engine and uses models (stubs) for the Linux kernel's libbpf API. This allows exploring all program paths without path explosion from the real API implementation. eBPF-SE was previously part of PIX, a tool that extracts performance interfaces from network function code.

George Candea

Dependable Systems Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Code AnalysisDebugNetwork
PapersArticles
Source Code

TPOT

Practical verification of system-software components written in standard C

Artifact repository for the paper "Practical Verification of System-Software Components Written in Standard C" (SOSP '24). Provides tools and methodology for verifying system-software components written in standard C.

George Candea

Dependable Systems Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Code AnalysisLow-Level
PapersArticles
Source Code

Twins

Linking patients to their physical and digital TWINs

Twins links patients to their physical biobank samples (organoids) and digital health records using blockchain-based access control and encrypted identifiers. The system ensures that researchers can access necessary biological or clinical data without being able to re-identify the contributing patient, combining OmniLedger smart contracts with privacy-preserving credential management.

Bryan Ford

Decentralized Distributed Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Access ControlConsensus
PapersArticles
Project Homepage
Contact

ProShare

Privacy-preserving photo sharing

Proshare allows you to share your photos, while protecting sensitive regions and hiding secret messages.

Touradj Ebrahimi

Multimedia Signal Processing Group

Maturity
Support
C4DT
Inactive
Lab
Unknown
Images
PapersArticles
Project Homepage

AI Research Aiding Online Children Safety

Platform for detecting specific risks and threats children are facing online

Machine Learning classifiers and tools to help with the detection of specific risks and threats children are facing while using their newly acquired phones. The classifiers perform on par or better than what exists on the market, and have been adapted to run smoothly on a mobile phone. The analysis can therefore run on the mobile phone without having to send any content to any server.

Karl Aberer

Distributed Information Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
Image Classification
PapersArticles
Project Homepage
Source Code
Contact

AI-Driven Classifier Building Pipeline

Platform for speeding up the development of Artificial-Intelligence classifiers

The platform aims at accelerating the process of building Artificial Intelligence classifiers for image and text classification.

Karl Aberer

Distributed Information Systems Laboratory

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
Image Classification
PapersArticles
Project Homepage
Source Code
Contact

CommPass

News article similarity visualization tool

The project aims to visualize semantic similarity and polarity of news articles covering a given media event. It extends the NewsTeller platform which processes over 1.5 million news articles per month. The goal is to provide intuitive visualization tools to explore the media landscape and enhance the NewsTeller platform.

Karl Aberer

Distributed Information Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
AdversarialMachine LearningNatural Language
PapersArticles
Project Homepage

Upgrading the Newsroom

Automated image selection system for news articles

The system fuses multiple textual sources (caption, body, headline, lead) from news articles using a hierarchical attention mechanism to retrieve relevant images. It utilizes subword embeddings and self-attention to better encode entities and capture important keywords within texts. The model is trained on a large-scale multimodal multilingual dataset of over 500k German and French news article-image pairs in a weakly-supervised manner.

Karl Aberer

Distributed Information Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
AdversarialMachine LearningNatural Language
PapersArticles
Project Homepage
Contact

cloud-performance-monitor

Monitoring and improving performance of cloud services

The project on machine learning for HPC servers and data centers at the ESL lab focuses on utilizing machine learning-based approaches for multi-objective resource management. This research line aims to enhance the efficiency of heterogeneous High Performance Computing (HPC) servers and data centers by employing system-level resource management techniques.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
BenchmarkCloudOptimization
PapersArticles
Project Homepage

digital-twins

Virtual copies for real-time connections

Digital Twins involve creating virtual copies of real objects for real-time monitoring and analysis. These replicas are utilized in various fields like disease prediction, climate policy assessment, and resource conservation. EPFL's DIGIPREDICT consortium used digital twins to forecast COVID-19 progression in patients.

David Atienza

Embedded Systems Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
AnonymityMedicalPredictor
PapersArticles
Project Homepage

carbon

Blockchain using lightweight broadcast-based primitives instead of consensus

Carbon is the next iteration on AT2, adding many new features, such as minting and a dynamic network. By decoupling leader selection from transaction processing, this system is more efficient and more scalable compared with Bitcoin and consensus-based solutions.

Rachid Guerraoui

Distributed Computing Lab

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
Byzantine ResilienceConsensusDecentralized
PapersArticles

Who is that?

Machine Learning framework to search a database via similarity samples

Use machine learning to organize a collection of objects (e.g. faces of actors) so as to make them interactively searchable via similarity queries. The approach combines concepts from representation learning, reinforcement learning, discrete-choice theory and bayesian inference.

Matthias Grossglauser

Computer Communications and Applications Laboratory 4

Maturity
Support
C4DT
Inactive
Lab
Unknown
Database
PapersArticles
Project Homepage
Contact

Predikon

Predictions about votes in Switzerland

We designed a predictive model of the municipal results (in percentage of "yes") for Swiss votes. Based on past results, our model automatically learns features of municipalities. On the day of a vote, and as soon as some municipalities publish their results (their percentage of "yes"), our model uses the features to predict the unpublished results of other municipalities. The prediction of the national vote result is obtained by aggregating published results and predicted results.

Matthias Grossglauser

Computer Communications and Applications Laboratory 4

Maturity
Support
C4DT
Inactive
Lab
Unknown
Predictor
PapersArticles
Project Homepage
Contact

Climpact

Estimate your perception of carbon footprint

Climpact uses pairwise comparison methodology and discrete-choice models to measure and analyze people’s perception of carbon footprints across various everyday actions. By aggregating crowd-sourced judgments, it quantifies systematic biases between perceived and actual CO2 impact, producing calibrated estimates useful for climate communication research.

Matthias Grossglauser

Computer Communications and Applications Laboratory 4

Maturity
Support
C4DT
Inactive
Lab
Unknown
Predictor
PapersArticles
Project Homepage
Contact

SLICO Superpixels

Efficient Superpixels

SLICO is a zero-parameter variant of the SLIC superpixel algorithm that adaptively normalizes gradient magnitudes across the image, producing compact superpixels of uniform size in both textured and texture-less regions. It runs efficiently in O(N) time and requires no manual tuning of the compactness parameter.

Sabine Süsstrunk

Image and Visual Representation Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Computer Vision
PapersArticles
Project Homepage
Source Code

Image Completion

Complete an image whose 99% pixels are randomly missing

This work addresses extreme image completion where 99% of pixels are randomly absent. The method formulates the problem as low-rank matrix recovery, exploiting non-local self-similarity and sparsity priors to reconstruct plausible image content from the tiny fraction of observed pixels, outperforming prior inpainting approaches under this highly degraded setting.

Sabine Süsstrunk

Image and Visual Representation Lab

Maturity
Support
C4DT
Inactive
Lab
Unknown
Computer Vision
PapersArticles
Project Homepage
Source Code

Food & You

Track your individual response to food and help advance science

Food & You is a digital epidemiology cohort study that combines continuous glucose monitoring (CGM), dietary logging via the MyFoodRepo app, and gut microbiome sampling to investigate personalized glycemic responses to food. Participants scan barcodes and photograph meals, generating longitudinal data used in downstream nutritional and microbiome research.

Marcel Salathé

Salathe Lab

Maturity
PrototypeIntermediateMature
Support
C4DT
Inactive
Lab
Unknown
MedicalSurvey
PapersArticles
Project Homepage
Contact

Tandem / Monero

Applying the privacy-preserving algorithms of Tandem to the usage of the Monero blockchain.

One challenge when using private keys is that if they get stolen the damage can be very big, especially when used in blockchains to protect funds. Sharing the private key between multiple entities can secure the usage a bit, but most algorithms don't give privacy. In the context of a privacy-preserving service like Monero, where the source and destination of a transaction are hidden, these algorithms would leak information about the user. Applying Tandem to transactions on the Monero blockchain allows to protect the private key in case of loss, while keeping the owner of the transaction private.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
AnonymityProtection
PapersArticles

Private Collection Matching Protocols

Framework to build privacy-preserving Private Collection Matching (PCM) systems

We introduce Private Collection Matching (PCM) problems, in which a client aims to determine whether a collection of sets owned by a server matches their interests. Existing privacy-preserving cryptographic primitives cannot solve PCM problems efficiently without harming privacy. We propose a modular framework that enables designers to build privacy-preserving PCM systems that output one bit: whether a collection of server sets matches the client's set. The communication cost of our protocols scales linearly with the size of the client's set and is independent of the number of server elements. We demonstrate the potential of our framework by designing and implementing novel solutions for two real-world PCM problems: determining whether a dataset has chemical compounds of interest, and determining whether a document collection has relevant documents. Our evaluation shows that we offer a privacy gain with respect to existing works at a reasonable communication and computation cost.

Carmela Troncoso

Security and Privacy Engineering Laboratory

Maturity
Support
C4DT
Inactive
Lab
Unknown
ProtectionProtocol
PapersArticles

DECOPT

Decomposition Convex Optimization

DECOPT is a MATLAB package for solving generic constrained convex optimization problems using a primal-dual algorithmic framework based on model-based excessive gap methods. It handles composite objective functions with arbitrary convex constraints and is backed by convergence guarantees proven in the accompanying NeurIPS 2014 and technical papers.

Volkan Cevher

Laboratory for Information and Inference Systems

Maturity
Support
C4DT
Inactive
Lab
Unknown
Optimization
PapersArticles
Project Homepage
Source Code

LB-CSMRI

Magnetic imaging resonance optimization

LB-CSMRI is a MATLAB framework that formulates MRI reconstruction as a learning-based compressed sensing problem. It jointly optimizes undersampling patterns and reconstruction parameters using a bilevel optimization strategy, enabling high-quality image recovery from fewer measurements than traditional methods require.

Volkan Cevher

Laboratory for Information and Inference Systems

Maturity
Support
C4DT
Inactive
Lab
Unknown
ImagesMedicalOptimization
PapersArticles
Project Homepage
Source Code