With an estimated billion surveillance cameras capturing people’s daily activities, chances are you’re regularly recorded if you live in a densely populated area.
In London, for example, there are said to be 500,000 CCTV cameras and the average person is estimated to be recorded 300 times per day.
To help people understand when they’re on camera and how ubiquitous cameras affect their privacy, computer scientists affiliated with the University of Jyvaskyla in Finland have released open-source software called cctv-exposure “for quantifying human exposure to CCTV cameras from a privacy perspective.”
The software, part of a broader initiative to allow people to make better decisions about their privacy, is described in a paper titled, “CCTV-Exposure: An open-source system for measuring user’s privacy exposure to mapped CCTV cameras based on geo-location.”
The paper, authored by Hannu Turtiainen, Andrei Costin, and Timo Hamalainen, is an extended version of research presented at the Business Modeling and Software Design: 12th International Symposium (BMSD 2022) in Fribourg, Switzerland. It describes the software as analogous to a Geiger counter for detecting harmful radiation.
“When compared to exposure to ‘harmful environments’ such as exposure to radiation, the CCTV-Exposure system is intended to act like a ‘CCTV dosage meter’ for travel activities of privacy-minded individuals,” the paper explains.
The software, written in Python 3, requires two input files: a global position system exchange (GPX) file containing the GPS coordinates of a person’s travels, and an XML file of camera location coordinates.
This camera location data is not yet available, beyond a test file for the city of Jyvaskyla.
That said, there is an ongoing effort by the team to collect together a broader set of coordinates, using computer vision algorithms to identify surveillance cameras in Google Street View images. The researchers anticipate providing an API that will supply this camera location data to users to CCTV-Exposure, making the code readily useful. It’s hoped that people can also submit their own reports of camera locations to be included in the database.
Right now, without this camera data, this project is very much one to watch, rather than one to get started with right away. Unless, of course, you in the meantime create your own camera location tables and use that with the code, which you’re all free to do and share with others.
Once software has the info it needs – your journey and details of surrounding CCTV – it calculates where the supplied route was exposed to a security camera. Its JSON-formatted output includes:
Identity: identity information for the processed file, track, and segment
Distance: total distance traveled in the segment, total exposure distance, average and mean distance to cameras (in GPX points)
Time (if applicable): average speed, total segment time, exposure time
Percentages: exposure per total distance, exposure time per total time (if applicable)
Per camera data: time, distance, and all camera data available (location, field-of-view, etc.)
Number of unique cameras
There’s also a Rust version of the code but it requires time-stamped GPX files due to parser limitations.
In a test of the software based on Jyvaskyla route data captured from researchers using Garmin devices, average CCTV exposure came to 12.5 percent based on distance and 15.1 percent based on time.
Andrei Costin, assistant professor at University of Jyvaskyla in Finland (JYU.FI) and co-founder/CEO of security firm binare.io, told The Register in a phone interview that the CCTV paper is one of five models he and his colleagues have developed to promote CCTV awareness.
“It comes from the need to understand how big the privacy invasion is, to quantify it,” he said.
Costin said there’s a lot of discussion about how many cameras there are in China, in the UK capital, and elsewhere, but this is based on hearsay, marketing pitches, and obscure methodologies that are not sound, scientific approaches.
This led to Costin and his colleagues developing a way to better define CCTV camera coverage. As mentioned above, their approach relies on computer vision and machine-learning to identify and geolocate CCTV cameras captured in Google Street View images and to calculate where the cameras can see.
This data can be fed into CCTV-Exposure to make it all work, and it’s hoped that this info will be provided via an online interface. There is no timeline for the availability of this, and the team asked for people to contact them if they can help financially or technically to make it happen.
“We developed the system based on Google Street View because it’s the biggest source of street view imagery,” explained Costin. “But our system also allows users to submit in real-time, taking a snapshot of a CCTV camera while their location is enabled and sending that to our servers.”
This helps keep camera location data current.
Costin said the group working on this project – which incorporates work from other research papers – is developing a web app to enable netizens to submit camera location updates. The researchers have been using the app internally though it’s not yet ready for public consumption.
It’s important, Costin said, for people to understand how pervasive this technology is and how much more so it’s likely to become.
“In the next couple of years – it’s not ‘if’ but ‘when’ – this technology will be complemented by facial recognition,” he explained, adding that this may be used to present people on the street with personalized ads.
“This is a really creepy reality,” said Costin, “so we’re trying to make privacy-enhancing tools, to at least give back some power to ordinary users.”
Costin’s primary concern, however, is making people aware, so they can make privacy-informed choices. This might mean choosing routes with fewer cameras, he said, or with more cameras, if personal safety is a concern.
“If you’re going to some remote location in an unknown area, you might try to go by the route with the most CCTV cameras in the hopes that if something happens to you, there will be some evidence,” he explained.
Costin added that there’s something like one camera for every eight people on the planet and he expects the camera count will get much larger. “I think that the growth rate of CCTV cameras is insane,” he said. ®