The latest research from Cornell University reveals artificial intelligence can steal valuable data such as passwords just by listening to what people type on their keyboards.
In the study, researchers trained an AI model to classify laptop keystrokes using a smartphone-integrated microphone, which studied the waveform, intensity, and timing of the keystrokes.
They also used an AI image classifier CoAtNet to conduct the research and trained the classifier on 36 keystrokes made on a Macbook Pro.
The results revealed that the technology was able to classify the text with 95% accuracy, "the highest accuracy seen without the use of a language model."
Additionally, when the classifier was trained on keystrokes using video communication platforms Zoom and Skype, it achieved an accuracy rate of 93% and 91.7%, respectively.
"Our results prove the practicality of these side channel attacks via off-the-shelf equipment and algorithms. We discuss a series of mitigation methods to protect users against these series of attacks," the research concluded.
Should We Be Worried?
Time and time again, AI companies have been warned of the risks their technology has on personal data.
While the high accuracy rate of the AI model is concerning, researchers believe there are certain mitigation techniques users can employ to tighten online security and prevent data breaching.
The paper suggested users should consider altering their typing styles and randomizing their account passwords. "With the success of language-based models, passwords containing full words may be at greater risk of attack," it wrote.
Other possible solutions include using biometric authentication systems, audio filtering software, and password managers that make logging into accounts safer and easier.
Edited by Nikola Djuric
-author-avatar_thumbnail.jpg){kind=avatar}