cwa-documentation
cwa-documentation copied to clipboard
corona-warn-app
Impact of shadowing underestimated
Where to find the issue
In https://github.com/corona-warn-app/cwa-documentation/blob/master/cwa-risk-assessment.md
Describe the issue
Even minor impediments, such as a person between the two smartphones or a signal-impeding smartphone case, can cause the distance to appear to be twice as large as it actually is.
This underestimates the impact of shadowing.
Suggested change
Even minor impediments, such as a person between the two smartphones or a signal-impeding smartphone case, can cause the distance to appear to be ten times as large as it actually is.
Indeed, a body sitting between phones represents around a factor of 10 in the error on the distance calculation. See here for various sources, for instance this one, which has the merit of being very visual:
https://www.youtube.com/watch?v=SAi24ctpyZQ
Internal Tracking ID: EXPOSUREAPP-1918
Funny to see my video popping up ;) I also have my doubts about distance measurement with Bluetooth, it has been discussed at length in this MIT webinar. It suggests that these errors could statistically cancel out if we had frequent independent measurement (100 times during 10 minutes, so very 6 seconds), but right now it seems to be every 5 minutes.
Statements like this from the risk assessment:
All exposures for a diagnosis key that lasted less than 10 minutes in total (regardless of how close the smartphones came during that time) or during which the smartphones were more than 8 meters (73 dB attenuation) apart on average (regardless of how long the exposure lasted) are discarded as harmless.
seem way to optimistic - we cannot put equal sign between 73 dB and 8 meters, it can be 1 meter as well, depending on where the person keeps the phone. Maybe it would average out over time, but if scanning is every 5 minutes for 2 seconds then there is no way that it will average out. If it was up to me I'd just notify everyone about exposure regardless of duration / signal strength until we have better tech available (UWB)
Hi pdehaye, Thank you for raising this issue. The team will get back to you soon. Best regards MC Your Corona Warn-App Open Source Team
@pdehaye
Even minor impediments, such as a person between the two smartphones ...
... signal-impeding smartphone case ...
Similar to have the smartphone in the back pocket?
I think, some impediments might be also impediments to the particles we are worried about. And some may exist because the persons are not aware of the technical limitations or the proper usage.
@keugens Unless you have your mouth on your back, I don't get your point ;-)
Sorry, maybe I missed your point: this factor shouldn't be 2 but 10. I can agree with that, but not much more.
... minor impediments, such as a person ...
Not agreed. This seems to be a larger impediment, no free space between the phones, does not match factor 10.
And, assuming the persons wearing their phones on the front-side of their bodies, it seems this impediment is also an impediment for the infectious particles. This additional signal attenuation is pointing at least in the right direction. Therefore it is not clear for me why you call this an error. The final goal is not distance measurement, but detecting situations with more or less infectious particles around.
Re. phone cases: for me, this is a matter of proper usage (like wearing the phone in the back pocket) and not about an error.
Re. error: for a normal free space situation I would roughly estimate an error of factor 2 for all (95%) phones using the app. To talk about the errors you mentioned, without talking about systematic errors, is misleading and does not inspire confidence in my honest opinion.
... distance calculation
It's a measurement, I would say. Which does not exclude parameters, formula and - unfortunately - massive systematic errors.
... ten times as large as it actually is.
... as it is measured by the phone under normal free space condition, I would say.
And due to systematic errors, it could be 20 times as large as it actually is.
@keugens My point was that this two should be replaced by ten. A person in the beam (either passing by or always in between because of how the phone is carried) is a difference of close to 20db, i.e. a factor of 10. Leaving it as is conveys the wrong impression about the possible error.
https://github.com/corona-warn-app/cwa-documentation/pull/435 was closed without being merged. I'm unsure how to proceed with this issue. I don't like the idea of leaving issues open fore 2 years without any movement on it, so either this should be fixed or the wont fix label should be applied and this issue closed.
