A multi-institutional team of researchers developed malware to extract caller information by screening vibration data from ear speakers recorded by a smartphone's accelerometers.
The researchers used two newer Android phones, whose motion-sensor data is retrievable without requiring users' consent.
The models' larger speakers also provided more caller information than older models, allowing a machine learning algorithm to infer 45% to 90% of the word regions from their accelerometer data.
The researchers learned their EarSpy malware could identify repeat callers with 91.6% accuracy, determine the speaker's gender with 98.6% accuracy, and identify spoken numbers from zero to nine with 56% accuracy.
Texas A&M University's Ahmed Tanvir Mahdad said attackers would have to conceal EarSpy within a downloadable application to pull off the exploit.
From Texas A&M Engineering News
View Full Article
Abstracts Copyright © 2023 SmithBucklin, Washington, D.C., USA