Beware! Hackers can now figure out your password using a keyboard with a 90 percent accuracy rate due to AI
The researchers that made the discovery emphasise that since there are so many microphones in close proximity to keyboards, this new method poses a greater risk to keyboard security.
artificial intelligence
Highlights
- An AI-powered hacking technique predicts passwords with over 90 percent accuracy
- The technique guesses passwords by listening to keyboard sounds
- Researchers replicate keystrokes in video chats using laptop microphones
In the realm of AI, a groundbreaking hacking technique has been unveiled, harnessing the power of AI to predict passwords with an astonishing accuracy of over 90 percent. This cyberattack method involves training an AI model to recognise the distinct sound profiles of various keys on a keyboard. The research, which is currently awaiting peer review, was shared as a preprint on arXiv.
Researchers at Durham University in the UK, under the direction of Joshua Harrison, used a smartphone's built-in microphone to eavesdrop on keystrokes coming from an Apple MacBook Pro. After processing the audio data, their AI model successfully reproduced the same keystrokes with an astounding 90 percent accuracy.
Testing the method on video calls
The researchers experimented during video chats to see how well the hacking technique worked in actual situations. The laptop's built-in microphone recorded keystrokes during a Zoom session, which the AI model was able to replicate with a 93 percent accuracy rate. Similarly, the model replicated keystrokes with about 92 percent accuracy when using Skype.
Threat reduction & consequences going forward
The researchers that made the discovery emphasise that since there are so many microphones in close proximity to keyboards, this new method poses a greater risk to keyboard security. The AI system learns and recognises distinct patterns in each user's input, taking into account things like sound, intensity, and timing of each press.
However, the AI model's applicability isn't universal across all keyboards. The researchers note that separate training is required for each keyboard, necessitating additional reference data to link specific keystrokes to their corresponding characters.
The report advises changing typing habits to combat this new danger, with touch typing dropping identification accuracy from 64 percent to 40 percent. The danger of such cyberattacks can be reduced by utilising biometric passwords like fingerprint or face scanning, as well as random passwords with different letter cases.
This report emphasises how cyber risks are always changing, underscoring the critical need for ongoing attention to protect private data in a society that is becoming more and more digital. To ensure that sensitive information is protected as AI develops, cybersecurity solutions must likewise evolve.
COMMENTS 0