Description

A novel cyber-security threat, termed the "Mic-E-Mouse" attack, takes advantage of high-speed optical sensors in standard computer mice to secretly record conversations. The sensors, intended for accurate tracking of movements, can be used to pick up minute acoustic vibrations transferred via work surfaces and desks when people talk. A research team has built a complicated pipeline using machine learning and signal processing in order to recover intelligible speech from vibrations. In spite of issues such as low-quality signals and non-uniform sampling, the attack has proven quite successful, with 80% accuracy in speaker identification and only 16.79% word error rate in human experiments. The Mic-E-Mouse attack is especially alarming as it aims at consumer-level mice that are readily available for less than $50. As the cost of producing high-performance input devices drops, the attack surface expands, opening up a wider opportunity for attackers. Vulnerable mice are often employed in open-source software, e.g., creative programs or games, where the data gathering process is not visible to users. As most games and applications include networking code, attackers can use this to steal gathered information secretly without being suspected. This vulnerability shows a novel form of audio surveillance, elevating common computer peripherals into intelligent eavesdropping devices. With the capability to record conversations quietly, this attack poses a significant privacy concern for both individuals and organizations. With more widespread use of such sensors in consumer products, the threat of exploitation increases, necessitating more secure hardware and software options to avoid such quiet exploitations.