Researchers have developed an artificial intelligence (AI) that can recognize passwords in the Metaverse based on hand movements. But the technology also carries risks.
Since the Corona pandemic at the latest, our everyday lives have increasingly shifted into the digital world. Conferences and appointments now often take place online using tools such as Google Meet, Microsoft Teams or Zoom. With the Metaverse, Meta, among others, wants to go one step further. Because then we move through virtual space as an avatar.
But such technical achievements also bring with them some risks. As recently demonstrated by the University of Chicago, participants in a virtual space can be easily spied on. The researchers demonstrated how easy it is to steal passwords. To do this, the team developed artificial intelligence.
AI guesses passwords via hand gestures with 98 percent accuracy
The system monitored conference participants in the VR work environment Meta Horizon Workrooms. The sessions lasted about ten to 15 minutes, during which participants went about their tasks in the virtual world. At the end of the experiment, the researchers found that the AI correctly memorized about 86 to 98 percent of all keystrokes for 13 out of 15 people.
This is problematic in that artificial intelligence could also record passwords and other sensitive information in this way. If people work together in the digital world and the microphones transmit keystrokes into the virtual world, cyber attacks are possible. The two remaining participants were protected from the attack by noise.
Passwords made from cryptic character strings are recommended
The easiest way for people in the metaverse to protect themselves is to use cryptic passwords. If a character string does not consist of words, it is more difficult for the algorithm to subsequently guess incorrectly recorded entries. This is much easier with a letter twist in a regular word.
Further research should show to what extent companies that offer VR applications can protect their user groups. Ultimately, with every further development, new scenarios for exploiting certain vulnerabilities arise. Researchers recently demonstrated that AI can guess passwords based on typing noises in video conferences.