Almost four weeks after the nation-wide launch of its centralized contact-tracing app StopCovid, the French government is patting itself on the back. And not least because StopCovid is one of the only European apps that both works, and also rejects the model put forward by tech giants Apple and Google.
Speaking at a press conference to update the public on the app's progress last week, the French secretary for the digital economy Cedric O, was keen to praise StopCovid for its "sovereign" quality.
"At a time of domination by the GAFA (Google, Apple, Facebook, Amazon)," said O, "it is not totally anecdotal that a French team managed to provide the health services with an app, before any of its European equivalents, all of which had to rely on an American infrastructure developed by Apple and Google."
"In the current global context, it seemed important to me to insist on the app's sovereignty," continued O.
SEE: Digital transformation: A cheat sheet (TechRepublic)
That was before O turned to the latest statistics, which unfortunately aren't all that impressive. Less than 2% of the French population has activated the app (1.8 million activations in total), and more importantly, a hefty 460,000 users have de-installed the tool.
Among StopCovid users, 68 people reported testing positive for the coronavirus, using the QR code that came with their test results. This let the health services trace back 205 interactions, and identify a grand total of 14 risky contacts, who were notified that they should get in touch with their GP.
O acknowledged that the numbers reflected a rather low up-take, but attributed it mostly to the low prevalence of the virus over the past few weeks. The number of COVID-19 cases has slowed markedly in France since the start of June, with the country now fully in the "green", meaning that restaurants, bars and cafes have re-opened.
In other words, don't blame the technology. "The app works well," said O. Although there is a caveat: "In a few words, we detect most of the contacts we should detect, with the detection limits that we know on iPhones," continued O. "We detect iPhones less well for the reasons that we all know."
Without putting an exact figure on the app's ability to detect iPhones, O assured that the work-around techniques developed by the app's engineers are having the "expected results".
Snubbing the technology developed by Apple and Google effectively comes at a price; in this case, the engineers who developed StopCovid had to find a way to make the app work on iOS and Android even without the tech giants' help.
Apple's devices are designed to reduce an app's functionality when it is running in the background. While the model the Cupertino giant developed in partnership with Google allows developers to by-pass the restriction, any other third-party app would by default have to be up on the screen to be fully functional.
Early on, the French government asked Apple to relax the Bluetooth standard, which hugely hampers the ability of an independently developed contact-tracing app to run effectively. When Apple refused to budge, O condemned the company's attitude, saying that its failure to cooperate would be something "to remember".
Despite the reassuring tone from O, it is still unclear how effective the StopCovid work-arounds really are. In the UK – one of the only European countries that followed France in initially snubbing Apple and Google's API – similar attempts to bypass the tech giants' rules delivered poor results.
Similar to the French app, NHSX's solution consisted of "waking up" the back-grounded app on iPhones by broadcasting a signal from Android phones. Although the concept worked in theory, the vast majority of iPhones still went undetected by the UK's app. NHSX eventually announced that it would work on a hybrid contact-tracing app that will integrate elements of Apple and Google's model.
In the UK, the NHSX revealed the contact-tracing app source code roughly at the same time as it launched trials for the technology on the Isle of Wight at the beginning of May. In France, parts of the StopCovid protocol were open sourced from mid-May, but the full source code was only made public at the start of June, when the technology was rolled out nationally.
Of course, some tests were carried out while the app was still in beta version, and a white-hat hacking campaign was launched to resolve security flaws ahead of the national release. When StopCovid launched, O said that the app had been tested successfully on the top 100 phone models that are most popular among French consumers.
SEE: Contact-tracing app: How did the UK go so badly wrong?
Jean-Marie Gorce, professor of telecommunications at the Insa Institute in Lyon, and who worked on the app's distance measurements and risk assessments, maintained that StopCovid has been effectively tested in controlled environments.
"We tested the app in different scenarios, particularly to make sure that we could estimate distance, and it worked well," Gorce told ZDNet. "We did it at speed, but that was due to our time frame. And the issue is also to have some benchmark information for these results."
Real-world data, therefore, is limited, and to an extent, the app's launch is a continuation of its test phase. O said that work had been continuous since the 2nd of June, when StopCovid became publicly available, and that engineers were still analyzing user feedback to work on future iterations of the technology.
Anecdotal evidence shows that StopCovid has been received with mixed feelings – and the high number of app de-installs reflects that. Users told ZDNet that they had deleted the app because it doesn't seem useful enough: "There have only been 14 notifications sent so far," said one user, "and the app sometimes de-activates all by itself." Other users said that they had deleted the app after it was reported that StopCovid would collect more personal information than originally planned.
The app, as a centralized tool, was always intended to register anonymized encounters between users in a central database, to let public health services run analytics on the propagation of the epidemic.
But shortly after the technology launched, Gaetan Leurent, researcher at the Inria Institute found that StopCovid was in fact more data-greedy than previously thought.
Leurent explained that the original version of StopCovid sent every single Bluetooth interaction to the central database, for the server to figure out which interactions had happened for long enough and at a close enough distance to be epidemiologically problematic.
"A better thing to do in terms of data minimization would be to do that evaluation on the phone itself, so you don't have to register every single encounter," said Leurent.
Acknowledging the issue, O has since confirmed that a filter would be in place in the new versions of the app, to make sure that contacts that happened for too little time, or at too far a distance, wouldn't be recorded in the central database.
SEE: Coronavirus: Business and technology in a pandemic
It remains that a public less trusting of the technology is not likely to boost the underwhelming download numbers. In contrast, the German contact-tracing app, which was released a few weeks after StopCovid and is based on the decentralized protocol developed by Apple and Google, has been downloaded 12 million times.
Without defending any particular model, Leurent argued that the privacy issues identified with the app will be a source of on-going concern. "When you introduce a new tool, and especially a medical tool, you have to do some evaluation of the risks and some clinical trials to see what good and what harms it will cause," he said.
"For the contact-tracing app, this hasn't been done. There is very little talk from the government about the risks of those applications, and that's concerning," he added.
Yet as it looks forward, the French government is laying out ambitious goals for StopCovid, including interoperability with other European apps, and even integrating the technology with other platforms, such as smart objects. But at the top of the list, unsurprisingly, health officials want to increase the app's uptake in anticipation of a second wave of infections.
Leurent, for his part, has reserved optimism for the future of the technology. "Of course, if the pandemic was more prevalent, there would be more usage," he says. "But I'm not sure whether that would be 5% uptake, or 50% uptake. As it is, it won't make much of a difference, and I wouldn't say it's a great success."