Exploring Practical Vulnerabilities of Machine Learning-Based Wireless Systems

Explore practical vulnerabilities of machine learning-based wireless systems in this 18-minute conference talk from NSDI '23. Delve into the design and evaluation of feasible adversarial attacks against ML-based wireless systems for communication and sensing applications. Learn about the unique challenges in the wireless domain, including lack of synchronization between benign and adversarial devices, and the effects of wireless channels on adversarial noise. Discover RAFA (RAdio Frequency Attack), the first hardware-implemented adversarial attack platform against ML-based wireless systems, and examine its impact on state-of-the-art communication and sensing approaches at the physical layer. Gain insights into the significant performance drops experienced by these systems in response to adversarial attacks, highlighting the importance of addressing vulnerabilities in ML-based wireless technologies.