This is an effort to have a broad and comparative discussion around risk scoring algorithms in Bluetooth based corona-tracking apps, highlighting subtle nuances in their conception and parametrization.
It is worth reminding all those going through those pages that:
- Bluetooth only serves in those apps as a proxy for distance;
- that it is very imperfect;
- that there is a lot of selection bias in the data;
- that it is not clear such interventions are more helpful than alternatives.
We currently use the following color coding, which will evolve:
Project | Authority | Protocol | Code Repository | Reproducible Build | Parameter Transparency | Parameter Update Transparency | COVID+ key repo | |
---|---|---|---|---|---|---|---|---|
StopCovid | #StopCovid | France | custom | GitHub | robert-server/#22 | Not applicable | ||
SwissCovid | @SwissCovid, #SwissCovid | @BAG_OFSP_UFSP, Switzerland | Google/ Apple Exposure Notification | GitHub | #3 | #13 | ||
ApturiCovid | @apturicovid, #ApturiCovid | Latvia | Google/ Apple Exposure Notification | security by obscurity | ||||
SmitteStop | @apturicovid, #SmitteStop | Statens Serum Institut, Denmark | Google/ Apple Exposure Notification | security by obscurity | ||||
Immuni | #ImmuniApp, #Immuni | Extraordinary Commissioner for the COVID-19 Emergency, Italy | Google/ Apple Exposure Notification | GitHub | #5, Android v1, iOS v1 | #14 | ||
ProteGo Safe | #ProteGoSafe | ??, Poland | Google/ Apple Exposure Notification | GitHub | #6 | #10 | ||
CoronaWarn | #CoronaWarnApp @CoronaWarnApp | Germany | Google/ Apple Exposure Notification | GitHub | #7 | #9 | ||
CovidTracker | Ireland | Google/ Apple Exposure Notification | GitHub | TBD | TBD | TBD | ||
StoppCorona | GitHub | Austria | See https://github.com/austrianredcross/stopp-corona-documentation/wiki/Key-matching-process#batches-provided-by-the-server |
Other projects analyzed:
-
UK: NHSx
-
Austria: StoppKorona
-
Norway: SmitteStopp
-
Netherlands: https://github.com/minvws
-
< many we are missing, listed in PersonalData.IO's database>
- expand the list above
- contribute an issue describing your understanding of how risk scoring works in the app
Please consider:
- Android vs iOS
- Bluetooth ranging (is there an explicit step to produce a distance measurement?)
- How is risk calculated from a distance measurement?
- How transparent the app is about the whole process?
- see also #2
- ...
If you have experience contributing to or setting up a project such as this one, please provide suggestions on how to improve (for instance which features should we use?)