A team of researchers from Pennsylvania State University has created a groundbreaking technology that utilizes millimeter-wave radar and artificial intelligence to eavesdrop on phone conversations from distances of up to three meters. This system can recognize and transcribe conversations with an accuracy of about 60% by analyzing the faint vibrations produced by the phone receiver during a call. This is reported by Interesting Engineering.

The research is based on adapting the open-source speech recognition model Whisper, which was retrained on radar data, enhancing the speech recognition quality despite the noisy and low-quality nature of the data. The radar was positioned approximately three meters away from the phone, and the AI was capable of recognizing up to 10,000 words.

Although the system's accuracy is not perfect, even partial recognition of key phrases could pose serious privacy concerns. Researchers liken their technology to lip-reading — while not recognizing every word, it helps to grasp the essence of the conversation, especially when considering context.

Importantly, this technology raises concerns about potential privacy threats, as malicious actors could exploit it for remote eavesdropping on private conversations without the user’s awareness. The researchers emphasize that their work aims to inform society about such risks and promote the development of protective measures against this type of spying.

The study is supported by the National Science Foundation, and the team plans to continue their efforts in safeguarding personal information from emerging technological threats.