true77
Hacking passwords by recording the sound of keystrokes is not a new concept, but recent research using AI has significantly improved its accuracy.
Researchers from Durham University, University of Surrey, and Royal Holloway University of London conducted a cyberattack simulation using a deep learning model that classified keystrokes based on audio recordings from Zoom and smartphone microphones. The results were astonishing, with 93 percent accuracy using Zoom recordings and 95 percent accuracy using smartphone recordings. This study showcased the potential of such attacks using readily available equipment and software.
The Threat of Acoustic Side Channel Attacks
This type of cyberattack, known as an acoustic side channel attack (ASCA), was initially explored in the early 2000s but has not been a major focus until now. With the increasing prevalence of video conferencing and remote work in public places like cafes, coupled with advancements in neural networks, the researchers warned of the rising threat of ASCAs. Their study, presented at the 2023 IEEE European Symposium on Security and Privacy Workshop, aimed to emphasize the viability of these attacks and raise awareness now that deep learning tools are more accessible.
Protecting Against ASCAs
To safeguard against acoustic side channel attacks, individuals are advised to avoid typing passwords near microphones or on platforms like Zoom whenever possible. However, for situations where this is unavoidable, researchers recommend implementing two-factor authentication, using biometric login options, and creating randomized passwords with a mix of upper and lower case characters to obscure the use of the Shift key. These precautions can help mitigate the risk of falling victim to ASCAs.
Topics:
– Cybersecurity
– Privacy
