Camera shake can identify you 4


Chris Baraniuk reports in New scientist magazine that footage from wearable body cameras contain a “motion signature”, unique to the user.

Shmuel Peleg and Yedid Hoshen at Israel’s Hebrew University of Jerusalem collected footage from 34 people who wore GoPro cameras on baseball caps. They ran it through an algorithm that recognised motion signatures particular to each person. The algorithm predicted the wearer with 88% accuracy and only required 12 seconds of video to make an identification.

This technique could be used to identify people who upload videos to sites such as Youtube, such as protestors uploading video from demonstrations. However, Peleg also pointed out that:

“On the other hand, if police officers have to wear cameras, this may give another level of assurance that the video you are being shown is from that officer and not someone else. It’s a double-edged sword.”

The original research paper can be found here.