AI Can Listen to Keyboard Sounds for Stealing Passwords

Initially, our concerns centered around webcams and microphones, but now our keyboards are also facing threats. Laptop users now face a potential risk of their confidential information, including personal messages, passwords, and credit card details, being compromised solely through their typing sounds.

A recent research paper, authored by a collaboration of scholars from British universities, highlights the remarkable ability of artificial intelligence to decipher keystrokes solely based on auditory cues, achieving an impressive accuracy rate of 95%. The evolution of AI suggests that these attacks will likely become more sophisticated in the rapidly advancing technological landscape.

How the Process Works

The research delves into the realm of “acoustic side channel attacks,” where malicious entities employ auxiliary devices like cell phones placed adjacent to laptops or unmuted microphones on platforms such as Zoom to capture the auditory input of typing sounds.

The captured sound is then processed through a deep-learning artificial intelligence model that has been trained to recognize distinct auditory patterns associated with individual keystrokes, ultimately revealing the typed text’s content.

Also Read: Islamabad’s Food Safety Standards Exposed in Operation

Through this method, researchers successfully identified keystrokes performed on a MacBook Pro solely by analyzing the associated sounds using a nearby mobile phone, achieving an impressive success rate of 95%. Similarly, analyzing a recorded Zoom call yielded a highly concerning identification accuracy of 93%.

The paper explains:

Considering recent advancements in microphone performance and the accessibility of deep learning models, the possibility of an acoustic attack on keyboards is increasingly plausible.

Protecting Against Such Threats

One approach to thwarting such attacks is to use stronger passwords that incorporate a mix of upper and lower-case letters, making it challenging for AI to decipher combined keystrokes. Employing passwords comprising a combination of numbers, letters, and special characters rather than complete words can also enhance security.

Additional security measures, such as implementing two-factor authentication and biometric verification, can provide added protection. However, the paper cautions that as AI continues to advance, it may potentially overcome various other security safeguards in the future.