Concerning user privacy in Meta’s Virtual Reality (VR) ecosystem, the Metaverse, new research has revealed some unsettling results.
The largest VR study of its kind was conducted by graduate researcher Vivek Nair and his team at the University of California, Berkeley’s Center for Responsible Decentralized Intelligence. They examined user interactions with VR to ascertain the degree of privacy.
There have been numerous cameras and microphones built into VR devices that can recognize faces, voices, and the user’s surroundings. These devices have been the focus of previous studies on privacy and VR. Advances in brain scanning technology that can be incorporated into headsets are a future concern for privacy advocates.
The UC Berkeley study demonstrates that none of that is even required; all that is required is the movement data of the user’s head and hands.
While playing the VR game Beat Saber, which necessitates nearly constant hand and occasionally head movement, over 50,000 subjects were observed, with over 2.5 million VR data recordings associated with them.
Individuals could be uniquely identified with a startling 94 percent accuracy with just 100 seconds of motion data using cutting-edge AI analysis. Furthermore, more than half could be located with just two seconds’ worth of information.
Several strategies have been proposed to avoid the loss of user privacy in VR. One is to obfuscate the motion data before it is sent to outside servers. But doing so would mean adding noise, which might impair the accuracy with which VR headsets and controllers track user movement. This would be problematic for games like the aforementioned Beat Saber, which depend heavily on this.
Berkeley researchers are researching methods that could be used to protect user privacy by obscuring movement data that can be used to identify a user specifically without affecting the accuracy and efficiency of VR devices.