A Canadian built a DIY camera to surveil the police, and officers showed up at his home

Canadian YouTuber Anthony Sistilli built a DIY license plate recognition system modeled on Flock Safety cameras and set it to look for police cars. He showed the build in two short videos for his 140,000 subscribers, and on September 20 officers from Peel Regional Police in Ontario came to his home.

The system consists of a cheap camera from Amazon, a laptop, and machine learning software. According to Sistilli, he trained the model in about 20 minutes: it looks for cars resembling police vehicles and, if it is confident of a match, tries to read the plate. The phone sends GPS data to the laptop, and the photo, time, location, and plate go into a database that can be viewed on a map.

The conversation with the officers lasted about half an hour and mainly concerned what Sistilli plans to do with the data. The police said that tracking their cars could put officers at risk and help people committing crimes. By his impression, the officers were not familiar with many cases of controversial use of similar systems in the US.

"I pointed the camera the other way, started watching the watchers, and they turned out to have the same privacy concerns as we do," Sistilli says. He stresses that he did not and does not intend to build a shared map of all officers, but wanted to check how feasible it is at all.

Audience reaction

The support turned out to be mixed. Some people suggested making a shared map, but there were also those who talked on social media about violence and hunting down specific officers. Sistilli does not approve of that.

In the conversation with the police, he mentioned approaches used in the US: data retention policies and audit logs that do not disclose personal information or details of specific cases. In his view, he failed to get this difference across. Flock Safety cameras run on more than 120,000 devices in 49 states, and similar systems, according to Sistilli, are appearing in his Canadian area as well.

Image: Antti Leppänen CC BY 4.0 · License

Translation editors: Соловьев В., Пеков А.