Researchers at Penn State have unveiled a startling method to eavesdrop on phone calls from up to 10 feet away, combining millimeter-wave radar and AI to decode the minute vibrations emitted by a phone’s earpiece. This system, adapted from radar technology used in self-driving cars and 5G networks, measures microscopic surface movements caused by sound waves. By pairing this data with a modified version of the Whisper AI speech recognition model, the team achieved around 60% transcription accuracy for continuous speech—enough to potentially capture sensitive keywords. The work expands on a 2022 project where only single words could be detected, moving now into real-time conversation analysis.

The implications for privacy are profound. Even partial or imperfect transcripts, much like lip-reading, can be combined with context to infer entire conversations. While the researchers stress that their intent is to highlight vulnerabilities before bad actors exploit them, the technology underscores an unsettling reality: private conversations may be more exposed than we realize. Supported by the U.S. National Science Foundation, the study warns that in the wrong hands, such AI-radar combinations could become a powerful surveillance tool, pushing society into yet another front in the battle to safeguard personal privacy.

Stay Awake. Keep Watch.

SOURCE: Interesting Engineering