The researchers suggest several methods to mitigate this type of attack, including changing one's typing style, using randomized passwords with multiple cases, and using a second authentication factor. For sensitive information beyond passwords, they suggest playing fake keystroke sounds to mask the real ones, particularly during Skype and Zoom transmissions. Further research is being conducted into using new sources for recordings, better keystroke isolation techniques, and the addition of a language model to improve their acoustic snooping.
Key takeaways:
- Researchers in the UK have developed a method to translate the sound of laptop keystrokes into their corresponding letters with up to 95 percent accuracy using a nearby iPhone or through Zoom and Skype calls.
- The method uses deep learning and self-attention transformer layers to capture the sounds of typing and translate it into data, without relying on a language model.
- Defending against this type of attack can be done by changing one's typing style, using randomized passwords with multiple cases, and using a second authentication factor.
- The researchers also suggest playing fake keystroke sounds to mask the real ones, especially for sensitive information like company records or customer info.