They attacked a robotic vacuum cleaner and overheard what was going on in the room where it was working
Scientists from the United States and Singapore used a robotic vacuum cleaner to eavesdrop on the sound in the rooms and identify TV programs playing in the room where the vacuum cleaner was located. The performance is even more impressive as the Standalone vacuum cleaner are not equipped with a microphone. This work shows that any device with lidar technology can likely be used for eavesdropping.
We use these types of devices at home without thinking about it much. We have shown that although such devices do not have a microphone, we can rewrite their navigation system to eavesdrop on conversations and reveal confidential information, says Professor Nirupam Roy of the University of Maryland.
This in autonomous robots used Lidar system examines the environment with the help of lasers. Their light is reflected from the surroundings of the vacuum cleaner and fed into the sensors of the vacuum cleaner to create a room map. Experts have been speculating for some time that the maps created by autonomous vacuum cleaners, which are often stored in the cloud, can be used for advertising.
Image source: Pixabay
Mapping a room makes it possible to determine its size, i.e. the size of the entire apartment or house, from which conclusions can be drawn about the level of income or lifestyle. Roy and his team started thinking about how to get one with Lidar equipped device can be used to eavesdrop on noises in the rooms in which it is located.
Sound waves cause various objects to vibrate, and these vibrations create small changes in the light waves that reflect off of those objects. The feed in a vacuum cleaner uses light that bounces off uneven surfaces of varying densities. The vacuum cleaner's sensors only receive part of this reflected scattered light. So Roy and his team weren't sure if this piece of information was enough for eavesdropping.
First, however, the scientists hacked into the remotely autonomous robotto demonstrate that they can control the location of its lasers and transfer the data to their computer without affecting the vacuum cleaner's navigational skills. Once they did that, they experimented with two sources of sound. The first was a recording of a man reciting various numbers. The recording was played through computer speakers. The second source of sound was the television speakers, which played various programs. The scientists, on the other hand, intercepted a laser signal that was sent by the vacuum cleaner's navigation system and reflected off various objects near the sound sources. These items included a trash can, cardboard box, disposable grocery box, polypropylene bag, and items that we find on the floor.
The researchers then let the recorded signals through Deep learning algorithms who had previously been trained in recognizing human voice and identifying musical sequences from television broadcasts. It turned out that the system - LidarPhone - identified the spoken numbers with 90% accuracy and recognized the television programs played with over 90% accuracy.
The scientists emphasize that autonomous vacuum cleaners are just one of many examples of devices that use technologies like lidar. Similar attacks can potentially be e.g. B. against smartphone infrared systems for face recognition or infrared sensors for motion detection.