Experts Develop System to Block Eavesdropping on Smartphones

Researchers from Columbia University have introduced a new algorithm designed to block unauthorized eavesdropping through microphones on smartphones, voice assistants, and other connected devices. One of the key features of this innovation is its predictive capability. The algorithm anticipates what the user is about to say and then generates background noise in real time. This noise is intended to mask the user’s speech from any potential eavesdropper.

Currently, the system only works with the English language and has demonstrated an 80% success rate in blocking eavesdropping during tests. The noise level is carefully calibrated so as not to interfere with comfortable conversation between users. Additionally, the new system significantly hampers the effectiveness of speech recognition technologies, regardless of the software or microphone placement being used. According to the researchers, they plan to continue improving the algorithm and promise to add support for other languages in the future.

Testing and Results

The researchers tested their system using eight NVIDIA RTX 2080Ti graphics cards on hours-long speech recordings. According to their report, they found the optimal prediction window for the algorithm to be 0.5 seconds. Afterward, they challenged several speech recognition systems with the processed audio. In 80% of cases, these systems were unable to extract any words.