AI can determine personal information through AR, VR users' motion data, studies say

Gaming data was used help determine identifying information, UC Berkeley researchers explain

People who participate in augmented and virtual realities are reportedly sharing more information than previously understood through motion data, according to researchers at U.C. Berkeley. 

In two studies published earlier this year, led by the university, authors found users can be identified using just minutes of their head and hand movements. 

Such data, which is collected, can be used to infer dozens of related characteristics, like age and disability status. 

"Users are revealing way more information than they think. They're revealing it, not just to the device or application, but to all the other users. And there's very little that they can do to prevent that," Vivek Nair, the studies’ lead author and a Ph.D. student at Berkeley’s Department of Electrical Engineering and Computer Sciences, said in a release. "That makes it a particularly salient threat."


Sather Tower on the campus of U.C. Berkeley

Sather Tower on the campus of U.C. Berkeley on Thursday, June 22, 2023, in Berkeley, California. (Jason Armond/Los Angeles Times via Getty Images)

Both papers were completed through U.C. Berkeley's Center for Responsible, Decentralized Intelligence as a part of the center’s Metaverse security and privacy research effort.

One study was based on a dataset that is more than 100 times larger than those in past studies, and authors used open source data from more than 50,000 Beat Saber virtual reality gamers. The release said it demonstrated that body movements are "as singular and reliable an identifier as fingerprints."

Researchers trained a machine learning classification model on an individual player’s motion data for 5 minutes each, finding that the model could identify that user in just 10 and 100 seconds with 73% and 94% accuracy, respectively.

In the second paper, academics created an adversarial virtual reality game aimed at collecting as much data as possible from 50 participants in a lab setting in just 10 to 20 minutes. 


A Meta Oculus Quest 2 virtual reality (VR) headset

Preview of the Meta Store in Burlingame, California, on Wednesday, May 4, 2022. (David Paul Morris/Bloomberg via Getty Images)

Authors were able to accurately identify or infer more than 25 characteristics, including location, age and height. 

While, based on ethical considerations, the study did not attempt to ascertain other personal indicators like sexual or political preference, researchers believe these kinds of inferences may also be possible.

Next, Nair aims to research defensive technologies to protect users' privacy, raising concerns about bad actors using these worlds to steal identities or expose information.

"We've done an extensive job of proving that there is a privacy risk here and that it is a different kind of privacy risk than what we have seen on the web," he said. "These kinds of approaches for how to either transform the data or control who has access to it, that's going to be our main focus moving forward."

An aerial view of the U.C. Berkeley campus

The U.C. Berkeley campus is seen from this drone view in Berkeley, California, on Thursday, March 16, 2023. (Jane Tyska/Digital First Media/East Bay Times via Getty Images)

Privacy and security risks posed here are currently most relevant to gamers. People purchased nearly 10 million virtual reality headsets last year, according to U.C. Berkeley.

Nearly half of the participants in both studies used Meta Platforms Inc.'s Quest 2, as well as the Valve Index and other headsets. 

No comments:

Powered by Blogger.