Study Shows That AI Tools Can Steal Passwords By Identifying The Key Strokes Upto 95% Accuracy
The recent developments in Artificial Intelligence have not only taken some really advance strides but have also raised a lot of concerns. These concerns are mainly about the misuse of these technological advancements for personal gains at the cost of others’ lives and reputations. The most concerning of all is the breach of privacy. And very recently this apprehension took form in a way that troubled everyone. A study shows that Artificial Intelligence tools can now steal passwords by listening to the keystrokes while typing.
Researchers from the Durham, Surrey, and Royal Holloway universities, in a study, tried to train an Artificial Intelligence model to replicate the password by identifying the sound generated by the keystrokes while typing the password on a 2021 version of MacBook Pro. Surprisingly, the AI model could replicate the password with an accuracy of 95%. The IEEE European Symposium on Security and Privacy Workshops published this research.
According to the researchers, when this model was used and keystrokes were recorded on a nearby mobile phone, the Artificial Intelligence generated results with the highest accuracy without the use of a language model. When the AI model generated the sounds by listening to a smartphone (iPhone 13 mini) which was kept 17 cm away from the keyboard, it generated the password with 95% accuracy. To test accuracy, the researchers pressed the keys “varying in pressure and finger”. Similarly, when the same experiment was conducted via Zoom call to check the accuracy of identifying the sound of keystrokes on that call, the AI model depicted a 93% accuracy rate. The researchers also highlighted that passwords containing “full words may be at greater risk of attack”.
The research thus conducted is proof that Artificial Intelligence Tools could be a gateway for scammers and hackers to carry out cyber frauds easily. Internet users are at risk of a big cyber threat/crime known as “acoustic side-channel attack”- a type of cyber attack that exploits sounds emitted by computers or other devices.
This whole threat is not very surprising if we realize the rate at which AI is advancing. But the solution to this can only be customizing modern tools to either cope with AI or to avoid such blatant threats to privacy. There is a need for technology but adapting it in a way where there is an elimination of the potential threat of cyber crimes.