Montreal AI lab develops 'privacy-first' contact tracing app to track COVID-19 cases

Canadian researchers are using Bluetooth technology to develop an app to trace, and even predict, the risk of contracting the novel coronavirus. McGill law professor Richard Janda explains how the app would work.
A team of Canadian researchers has been developing an app that will rely on Bluetooth technology and machine learning to trace and predict the risk of contracting the novel coronavirus. (lzf/Shutterstock)

If you use the Bluetooth on your phone, so far it's probably been for the usual reasons, like connecting to a speaker, earbuds or your car's stereo.

But what if it could be used to trace — and even predict — the risk of contracting the novel coronavirus?

A team of Canadian researchers has been developing an app that will use machine learning to do just that. 

The team is directed by AI expert Yoshua Bengio, and includes researchers from Mila (Bengio's research institute), the Université de Montréal, McGill University, the University of Toronto and The Decision Lab, a behavioural science think tank.

The idea of using Bluetooth for contract tracing has been developed in Asia, and U.S. tech giants Apple and Google are also collaboratively working on it.

The Canadian version aims to fully protect privacy rights as a first priority, so none of the information the app gathers will be stored in a central location.

"There's no point at which an authority, however well-intentioned they may be, controls a database that has the geolocation of Canadians sitting on it," Richard Janda, a law professor at McGill University and part of the team designing the app, told Spark's Nora Young.

"We might want to trust health authorities with that information, but if that information leaks, or if that information gets in the hands of the police, or if it gets in the hands of other rogue authorities, we may not like the outcome for Canadian society," he added. 

So rather than being stored in a central location, the data will actually be stored on individual users' phones, he said. Additionally, encrypted data would be sent to a mix of servers. Each server would shuffle the time and place of contact. This amounts to very robust security and privacy protection, Janda said.

To hear more of the interview with Richard Janda, and hear how the soon-to-be released app will work, click the "listen" button at the top of the page.