A group of researchers from the David Ben-Gurion University of the Negev and the Weizmann Institute (Israel) have developed a technique () to reconstruct conversation and music in an enclosed space using passive analysis of the vibration of an electric lamp in a pendant light. An electro-optical sensor placed on the street was used as an analyzer, with the help of a telescope aimed at a lamp visible through the window. The experiment was carried out with 12-watt LED lamps and made it possible to organize eavesdropping from a distance of 25 meters.
The method is valid for a suspended lamp. Sound vibrations create air pressure drops, which cause micro-vibrations of a suspended object. Such microvibrations lead to light distortions at different angles due to the displacement of the glow plane, which can be captured using a sensitive electro-optical sensor and demodulated into sound. A telescope was used to capture the light flux and direct it to the sensor. The signal received from the sensor (Thorlabs PDA100A2 based on a photodiode) was converted into digital form using a 16-bit analog-to-digital converter ADC NI-9223.
The extraction of sound-related information from the general optical signal was carried out in several stages, including , normalization, noise suppression and frequency amplitude correction. A script for MATLAB was prepared for signal processing. The quality of sound restoration when taking parameters from a distance of 25 meters turned out to be sufficient for speech recognition through the Google Cloud Speech API and for determining the musical composition through the Shazam and SoundHound services.
At the same time, in the experiment, sound reproduction in the room was carried out at the maximum volume for the available speakers, i.e. the sound was much louder than normal speech. The LED lamp was also chosen not by chance, but as providing the highest signal-to-noise ratio (6.3 times higher than an incandescent lamp and 70 times higher than a fluorescent lamp). The researchers explained that the attack distance and sensitivity can be increased by using a larger telescope, a better sensor, and a 24 or 32-bit analog-to-digital converter (ADC), while the experiment was conducted using a handy telescope, a cheap sensor, and a 16-bit ADC. .
Unlike the previously proposed method,βwhich captures and analyzes vibrating objects in a room, such as a glass of water or a bag of chips, Lamphone enables real-time listening, while a visual microphone requires hours of intensive computing to recover a few seconds of speech.β . Unlike methods based on the use or as a microphone, Lamphone allows the attack to be carried out remotely, without the need to run malware on devices in the room. Unlike attacks using , Lamphone does not require illumination of the vibrating object and can be produced in passive mode.
Source: opennet.ru