Researchers from Pennsylvania State University have introduced a groundbreaking technique that utilizes millimeter-wave radar combined with artificial intelligence to listen to phone conversations from up to three meters away. This system can recognize and transcribe conversations with an accuracy of approximately 60% by analyzing minimal vibrations produced by the phone during a call. This information is reported by Interesting Engineering.
The core of the research involved adapting an open-source speech recognition model called Whisper, which was retrained using radar data, allowing for improved speech recognition despite the high noise levels. The radar was positioned about three meters away from the phone, and the data was processed by AI capable of recognizing up to 10,000 words.
While the system's accuracy is not perfect, even partial recognition of key phrases can pose serious privacy risks. The researchers liken their technology to lip-reading, which also cannot capture every word but can convey the essence of conversations, especially when contextual information is taken into account.
This technology raises concerns regarding potential privacy threats, as malicious actors could employ it for remote eavesdropping on personal discussions without users’ knowledge. The researchers emphasize that their work aims to raise public awareness about these risks and promote the development of protective measures against such forms of surveillance.
The research is supported by the National Science Foundation, and the team intends to continue their efforts to safeguard personal information against emerging technological threats.
