The paper suggests several defenses against these types of attacks, including changing typing style, using randomized passwords with multiple cases, adding randomly generated false keystrokes to the transmitted audio of video calls, and using biometric tools like fingerprint or face scanning instead of typed passwords. The researchers note that laptops, particularly those with uniform, non-modular keyboards, are more susceptible to having their keyboard recorded in quieter public areas.
Key takeaways:
- Researchers have developed a method to interpret remote keystrokes with up to 90% accuracy by recording the sound profiles of individual keys and training a deep learning model.
- The researchers used a 2021 MacBook Pro and an iPhone 13 mini to test their concept, achieving higher than 93% accuracy in both tests.
- The position of a key plays an important role in determining its audio profile, with most false-classifications tending to be only one or two keys away.
- The paper suggests several defenses against these kinds of attacks, including changing typing style, using randomized passwords with multiple cases, adding randomly generated false keystrokes to the transmitted audio of video calls, and using biometric tools instead of typed passwords.