DolphinAttacks (i.e., inaudible voice commands) modulate audible voices over ultrasounds to inject malicious commands silently into voice assistants and manipulate controlled systems (e.g., doors or smart speakers). In this paper, we design a lightweight method that can not only detect such attacks but also identify the direction of attackers without requiring any extra hardware or hardware modification. Based on the physical properties that the inaudible voice commands attenuate faster than the audible ones, we design a microphone array with five microphones to capture the attenuation differences mainly caused by obstacles (surface of devices) and use the differences to detect attacks. we conducted experiments to evaluate the effectiveness of our method in terms of various factors, including carrier frequencies, attack distances, angles, background noise, and voice command types, etc. Our experiments show that our method can detect inaudible voice commands with an accuracy of 99% and recognize the direction of the attackers with an accuracy of 97.89%.
Guoming Zhang received his B.S. degree in mechanical engineering from Ludong University, Yantai, China, in 2013. He received his M.S. degree in mechanical and electronic Engineering from Beijing Institute of Technology, in 2016, and is currently working toward the PH.D. degree in electrical engineering at Zhejiang University, Hangzhou, China. His research interests are in the areas of IoT security, acoustic communication, especially for the security of speech recognition system.
Xiaoyu Ji is an associate professor with the department of Electrical Engineering of Zhejiang University. He received his B.S. degree in Electronic Information & Technology and Instrumentation Science from Zhejiang University, Hangzhou, China, in 2010. He received his Ph.D. degree in department of Computer Science from Hong Kong University of Science and Technology in 2015. From 2015 to 2016, he was a researcher at Huawei Future Networking Theory Lab in Hong Kong. His research interests include IoT security, wireless communication protocol design, especially with cross-layer techniques. He won the best paper awards of ACM CCS 2017, ACM ASIACCS 2018 and IEEE Trustcom 2014. He is a member of IEEE. The publications are available at the homepage.