Want to stay anonymous Don't wear a camera

15.12.2014
Next time you strap on a GoPro or other wearable camera, keep in mind that your movement pattern could someday be identified like a fingerprint.

A recent study out of The Hebrew University of Jerusalem (via The Verge) found that wearable camera users generate a biometric movement pattern in the recording as they walk. A computer algorithm could then pick up on this pattern with just a few seconds of video.

Much like a fingerprint, a single video alone couldn't reveal someone's identity. But it could be linked to other videos that contain personally identifiable information.

Why this matters: As the study's authors, Yedid Hoshen and Shmuel Peleg, point out, the ability to identify someone from a first-person video could be both good and bad. A device like Google Glass, for instance, could use movement as a form of authentication, allowing access to sensitive files or work sites for only a specific wearer. On the downside, a malicious advertiser could tap into this data for unauthorized user tracking, and oppressive governments could have an easier time identifying videos from protestors.

Don't slow down

"The implication of our work is that users' head-worn egocentric videos give much information away," the study's authors, Yedid Hoshen and Shmuel Peleg, wrote. "... Care should therefore be taken when sharing such raw video."

It's a scary thought, but users can at least take comfort knowing that the algorithm relied solely on the walking portions of test subjects' videos. It's unclear whether the same biometric data could be gleaned from sitting, running or other activities, so GoPro users recording their latest skiing and cliff-jumping exploits may be in the clear.

And while the authors note that 2D video stabilization didn't significantly affect the results, more elaborate stabilization methods could make identification more difficult. Tech companies are racing to smooth out the shakes in first-person video anyway, so perhaps this issue could resolve itself over time.

(www.pcworld.com)

Jared Newman

Zur Startseite