Practical Defense Against Adversarial WiFi Sensing

WiFi-based sensing, a non-intrusive technology that leverages existing communication infrastructure, has become widely used for environmental monitoring by extracting Channel State Information (CSI). However, the vulnerability of these systems to adversarial attacks, a core challenge within Integrated Sensing and Communication (ISAC), highlights the need for practical and robust defenses. In this work, we introduce a practical black-box defense strategy designed to protect CSI data from adversarial manipulation, significantly reducing an attacker’s classification accuracy from 98% to 17% while preserving communication quality. Our approach achieves a minimal median Signal-to-Noise Ratio (SNR) difference of 1 dB, ensuring stable throughput and reliable system performance. This defense represents a crucial step forward in securing WiFi-based sensing systems, offering a resilient, low-impact solution that safeguards both sensing integrity and communication efficacy.