A Irish study shows that the Corona warning app in trams does not reliably detect when the minimum distance to other app users is not reached. In no single case was a proximity alarm triggered when the app’s detection rules were applied. For example, the Corona warning app in public transport goes by the rags of many cases in which one person has come too close to another for too long.
Study: Guessing is just as accurate
Don’t miss any news! With our daily newsletter you will receive all the news from heise online for the past 24 hours every morning.
Subscribe to the newsletter now
The study from June of this year found that in the tram there is only a slight correlation between the received signal strength of the Bluetooth signals used for contact tracking and the distance values calculated from them. In other words: The app cannot determine from the signal strength in the train how far away the sending smartphone is. How the Corona warning app detects contact cases is described in detail in another article.
According to the study, there was no distance alarm when applying the German as well as the Swiss rules for their respective warning apps. When using the Italian recognition rules, there were 50 percent correctly recognized cases, but also 50 percent false alarms. The hard conclusion of the research group led by Douglas J. Leith and Stephen Farrell from the School of Computer Science and Statistics at Trinity College in Dublin: A random selection from the passengers would have resulted in a similar hit rate, without any approximation data.
The same smartphones for comparable measurements
For the study, the researchers recruited seven participants and gave them each a Google Pixel 2 smartphone. They then took up various positions in the tram and hired them out for similar activities, all of which ran over a uniform WiFi hotspot. So the parameters should be made controllable, uniform and comprehensible, especially for the differentiation of the distance.
The researchers used the Google / Apple Exposure Notification (GAEN) API as a basis for the following measurements and applied an exposure notification app modified by the researchers from Google to the detection rules of the examined apps.
For the test protocol, the researchers made several fifteen-minute rounds with the test subjects in fixed positions. In each of these passes, the GAEN-API made three scans. After that, the positions and other test conditions were changed.
High frequency goes strange ways
The reflections of the high-frequency Bluetooth signals on the metal structures of the trams resulted in considerable signal fluctuations when smartphones were kept unchanged: According to this, there were fluctuations of more than 10 decibels up and down. For the sake of simplicity, the researchers calculated with attenuation values, which they calculated back from the Bluetooth signal strengths.
The same can be expected in buses and other means of transport with a similar structure. Anyone dealing with high frequency will hardly be surprised by the results. Radio operators have long been saying: “High frequency goes strange ways!” This is also shown by the values.
As expected, the signal strengths were reduced between 0.5 and 1.5 meters away, so a greater distance meant greater attenuation. But then the attenuation level remained constant up to a distance of 2.5 meters. There was a sharp increase in attenuation again at 3 meters, which the researchers attributed to the special arrangement of the seating groups on the train – here metal elements apparently acted as reflectors. Regardless of this, one would normally expect a further increase in attenuation with increasing distance, but the attenuation actually decreased again. Overall, the attenuation at a distance of 1.5 meters was around 52 decibels, at 4 meters it was 60 decibels, with a very strange course of the attenuation values.
The measurements of the attenuation values fluctuated tremendously even with the same distance. At a fixed distance of two meters, for example, both 52 decibels and almost 80 decibels were measured.
Corona warning apps only save a contact when a short distance and long contact time come together. To keep the minimum distance below the minimum, the apps use several trigger thresholds, which must also be exceeded over a certain period of time before a contact ID is marked as possibly relevant for the spread of an infection and saved on the smartphone. Attenuation values jumping back and forth counteract this detection and make it difficult for the app to actually detect a contact.
The researchers’ findings will now be examined by others. For many people who come close to strangers on public transport, the results of other groups and the reaction of the app developers to them are likely to be an important decision criterion for further use of the app.
The public transport operators should also be interested in clarification. The very inadequate enforcement of mask requirements by the operators, exacerbated by the lack of opportunities to maintain minimum distances, has already led many people to use other means of transport, whether bicycle or car. But that is not practical for everyone.