Researchers Train AI to Steal Data From Keyboard Clicks

Randomized passwords could help protect against this new danger
By Steve Huff,  Newser Staff
Posted Aug 12, 2023 8:30 AM CDT
Researchers Train AI to Steal Data From Keyboard Clicks
A woman types on a keyboard in New York.   (AP Photo/Jenny Kane, File)

Debates over whether advancements in artificial intelligence are good or bad won't subside any time soon—and this new wrinkle won't help. British researchers have discovered that by mixing machine learning and recordings of typing, it's possible to train AI to steal passwords merely by hearing which keys get pressed. Via Tom's Guide and BleepingComputer comes the story of a so-called acoustic attack that can siphon data from keystroke sounds with remarkable 95% accuracy. According to the paper published August 3, the researchers positioned an iPhone a short distance from a MacBook and recorded the sound of 36 keys as each was pressed 25 times. The recordings allowed the researchers to make waveforms and spectrograms that enabled them to visualize the acoustic differences between each key.

Those images were used to train a deep-learning image classification program called CoAtNet, basically integrating visual recognition with sound to steal information. In theory, hackers could record keystrokes using a smartphone infected with malware or, nearly as successfully, via Zoom (93% accuracy) or Skype (91% accuracy). Basic protections for all devices include working in isolated environments—one University of Virginia paper makes it clear researchers were aware of this necessity nearly 50 years ago—and controlling microphone access in general. In their paper, the British researchers state that "simple typing style changes could be sufficient to avoid attack," and they recommend other measures such as using white noise and randomized "passwords featuring multiple cases." (More hacking stories.)

Get the news faster.
Tap to install our app.
X
Install the Newser News app
in two easy steps:
1. Tap in your navigation bar.
2. Tap to Add to Home Screen.

X