We do not host this demo anymore, but you can find its code here.
Together with our SwissRe partner we created a demonstrator using the Drynx data analysis tool. The source data is a list of house insurances and the incurred claims. Supposing that this list is spread over many insurers, Drynx allows to:
Looking at the start screen of Drynx below, you can see the 3 data sources (A), (B), and (C). For the demo, the data is visible to the user, but in the application itself, the data from one insurer is not available in clear to the other insurers.
Now you can click on (1) and choose the columns that you want to analyze. You can choose for
example the Design year and the Incurred claim to see how much a house from a certain
year will cost to the insurer. Of course this supposes that the houses themselves have the same value
over the years. This query is also possible, but left as en exercise to the reader!
In (2) you can choose the type of operation, but for multi-column analysis, only
linear regression is available.
Clicking on (3) will start the calculation. The nodes holding the data of the insurers (the Data Providers) will communicate their encrypted values and do the calculation only on these values. Once the result is calculated, it is sent encrypted to the browser, who can then decrypt it and display it. The resulting linear regression can be seen in the following screenshot:
You can see that older houses tend to have higher claims than newer houses, which seems intuitively correct.
Presentation
Demo
Technical