Shock.. Radar Eavesdrops on Your Phone via Internal Vibrations

Sada News - Researchers at the University of Pennsylvania have discovered that the vibrations of the earphone inside the phone - which activate when the phone is placed on the ear - can be captured by a radar sensor from a distance and converted into sound, and replicated using artificial intelligence, according to what was published by the New Atlas website.

Fortunately, the new technology, called WirelessTap, has not yet reached an ideal stage. It currently achieves a maximum accuracy of about 60% in recognizing words, works up to a range of 3 meters, and the accuracy decreases even within this range. However, this technology provides a glimpse into the future of espionage and surveillance, showing how artificial intelligence can prove its importance in these fields.

Millimeter Waves

The WirelessTap system relies on mmWave sensing technology, where radar sensors can capture radio frequencies in the range of 77-81 GHz for various applications, including detecting movement inside homes or vehicles, and assisting self-driving vehicles in detecting surrounding obstacles.

The researchers demonstrated how commercially available mmWave radar sensors can capture the nearly imperceptible vibrations of the phone's earphone and the vibrations generated by the phone itself from a distance of a few feet.

Computer scientist Suryodai Basak, a researcher in the study, who published it with a colleague in the Proceedings of WiSec journal, explains that if "these same vibrations are captured using remote radars and machine learning is employed to help understand what is being said, using contextual clues, entire conversations could be identified." He adds that "by understanding the available possibilities, the public can be helped to recognize potential risks."

Noisy Data

The researchers transformed the noisy data captured by the sensor into recognizable speech using a customized version of OpenAI's Whisper speech recognition and transcription model. While WirelessTap demonstrated an accuracy of about 60% at a close distance of approximately 50 cm, the researchers indicate that this could be improved by incorporating elements from the context of the conversation in question - similar to how lip readers can interpret what people they are monitoring are saying more accurately when they know what their targets are talking about.

Eavesdropping Techniques

The researchers point out that this system surpasses other radar-based eavesdropping techniques, which either relied on the victim using a speaker or were limited to distinguishing sound using a limited vocabulary. Thus, WirelessTap is more discreet.

The mmWave sensor can be miniaturized to fit common small objects like pens. With the contextual background of audio recording, WirelessTap could be an effective tool for espionage or other covert methods of stealing private information.

Sensitive Calls

Researcher Basak concluded by saying: "The aim of the study was to explore the potential use of these tools by misleading entities to eavesdrop on phone conversations remotely," adding that the results indicate "that this is technically possible under certain conditions, and it is hoped that the findings will contribute to raising public awareness, so that people exercise more caution during sensitive calls."