- Pivid banks on a statistical technique called differential privacy to gather and share individuals’ information with securing their privacy.
- A new surveillance video analytics system will help track the number of people passing and calculate the average car speed of passing cars.
A team of scholars has developed “Pivid,” a new system that allows video analytics in a privacy-preserving manner to tackle invasive tracking concerns.
Frank Cangialosi, the lead author of the study and a researcher at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), said, “We’re at a stage right now where cameras are practically ubiquitous. If there’s a camera on every street corner, every place you go, and if someone could process all of those videos in aggregate, you can imagine that entity building a very precise timeline of when and where a person has gone. People are already worried about location privacy with GPS — video data in aggregate could capture your location history and moods, behaviors, and more at each location.”
Differential privacy, a statistical technique that facilitates the collection and sharing of aggregate data while safeguarding individual privacy, forms the base of Pivid’s development.
The same is accomplished by adding random noise to the results to avoid re-identification attacks. The amount of noise added is a trade-off – more noise makes the data more anonymous and makes the data less useful – and it’s determined by the privacy budget. This ensures that the results are still correct and configured low enough to avoid data leakage simultaneously.
The querying framework includes a “duration-based privacy” approach, where the target video is temporarily divided into chunks of the same duration. Later, these chunks are uploaded separately into the analyst’s video processing module to generate the noisy aggregate result.
The idea behind this is that by adding specialized types of noise to the data or analysis methods, relevant parties can be prevented from identifying an individual while simultaneously not concealing the societal patterns findings that emerge while analyzing the performance on the video inputs like the number of people who passed by the surveillance camera on a particular day or calculating the average car speed observed.
It is also a perfect way to prevent malicious actors from singling out particular people and determining their presence or absence in the video.
“In building Privid, we do not advocate for increasing public video surveillance and analysis. Instead, we observe that it is already prevalent and is driven by strong economic and public safety incentives. Consequently, it is undeniable that the analysis of public videos will continue. Thus, it is paramount that we provide tools to improve the privacy landscape for such analytics,” Researchers said.