Is your vacuum cleaner listening to you? Researchers hacked a robot vacuum and recorded voice and music remotely-ScienceDaily

Researchers, including Nilpamroy, an assistant professor at the University of Maryland’s School of Computer Science, gathered information from the laser-based navigation system of popular vacuum robots and applied signal processing and deep learning techniques to recover voice. A program that is played in the same room as the device that identified the TV.

This study shows that devices using light detection and lidar technology can be manipulated to collect audio without a microphone. This work, a joint research project with Assistant Professor Jun Han of the University of Singapore, was presented at the Association for Computing Machinery’s Conference on Embedded Networked Sensor Systems (SenSys 2020) held on November 18, 2020.

“We welcome these devices to our home, and we’re not thinking about it,” said Roy, co-appointed at the University of Maryland Institute for Advanced Computer Science (UMIACS). Told. “But even if these devices don’t have a microphone, they can reuse the system they use for navigation to spy on conversations and reveal personal information.”

The lidar navigation system for home vacuum cleaner bots shines a laser beam around the room and senses the reflection of the laser as it hits nearby objects and bounces off. Robots use reflected signals to map rooms and avoid collisions as they move through the house.

Privacy experts say that maps created by vacuum cleaner bots, which are often stored in the cloud, cause potential privacy breaches, and advertisers can tell the size of the house (indicating income level) and other lifestyles. It suggests that it may give access to information such as related information. Roy and his team wondered if these robots, lidar, could pose a potential security risk as a recording device for users’ homes and businesses.

Sound waves vibrate an object, and these vibrations cause a slight change in the light that bounces off the object. Laser microphones, which have been used in espionage since the 1940s, can convert these fluctuations back into sound waves. However, laser microphones rely on a target laser beam that reflects off very smooth surfaces such as glass windows.

Vacuum riders, on the other hand, scan the environment with a laser to detect light scattered by objects of irregular shape and density. The scattered signal received by the vacuum sensor provides only a small portion of the information needed to recover the sound waves. Researchers were uncertain whether the vacuum bot’s lidar system could be manipulated to act as a microphone, and whether the signal could be interpreted as a meaningful audio signal.

First, researchers have shown that robot vacuums can be hacked to control the position of the laser beam and send detected data over Wi-Fi to laptops without interfering with device navigation.

Next, we conducted an experiment with two sound sources. One source was the human voice, which represents the numbers played on computer speakers, and the other was the sound of various television programs played on the TV soundbar. Roy and his colleagues captured the laser signals that the vacuum navigation system sensed as it bounced off various objects placed near the sound source. Objects include trash cans, cardboard boxes, takeaway containers, polypropylene bags, and more. These items are usually on the normal floor.

Researchers passed signals received through deep learning algorithms trained to match human voices or to identify music sequences from television programs. Their computer system, which they call the Lidar Phone, identified and collated the spoken numbers with 90% accuracy. It also identified TV programs from one minute of recording with an accuracy of 90% or higher.

“Given that we all order food over the phone and have meetings on the computer, this kind of threat can be more important than ever, speaking credit card and bank information. Often, “says Roy. “But what’s even more worrisome to me is that it can reveal far more personal information. This kind of information is my lifestyle, the number of hours I work, I do. You can tell you about other things that we are watching, and what we are watching can reveal our political direction on television. It controls political elections or I It’s important for anyone who might want to send a very specific message to. “

Researchers emphasize that vacuum cleaners are just one example of a potential vulnerability to lidar-based espionage. Many other devices can be exposed to similar attacks, such as smartphone infrared sensors used for face recognition and passive infrared sensors used for motion detection.

“I think this is an important task that will make manufacturers aware of these possibilities and help the security and privacy community come up with solutions to prevent these types of attacks,” Roy said. I will.

This research was partially supported by a grant from the Singapore Ministry of Education Academic Research Fund Tier 1 (Prize No. R-252-000-A26-133).

Research paper “Spying with a Robot Vacuum: Eavesdropping Through a Rider Sensor”, Sriram Sami, Yimin Dai, Sean Rui Xiang Tan, Nirupam Roy, Jun Han, November 18, 2020, at Association for Computing Machinery, SenSys 2020 It was announced. ..