An Adversarial Jamming Attack Detection Method Using Energy-Detection-Based Hardware for IoT Systems

This paper presents an adversarial jamming attack detection method using energy-detection-based hardware for wireless Internet of Things (IoT) systems. The proposed method utilizes accurate RF signal energy level detection to establish a learning baseline for identifying different types of jamming attacks in wireless networks, which are often characterized by changes in RF energy levels. An adaptive threshold-based binary classification algorithm is implemented to use the detected signal’s energy level as the classification metric. This enables the detection of deceptive/reactive and periodic jamming attacks. Energy levels of RF signal are measured using an ultra-low power (ULP) direct RF-to-digital Received Signal Strength Indicator (RSSI) circuit, developed in a 65-nm CMOS technology, which consumes only 6 nW of power. Further, the RSSI circuit incorporates a band-pass response to filter out ambient signal in the network, providing resilience against continuous RF jammers. Furthermore, a system-level model of the proposed method is presented to demonstrate its detection capabilities. To mitigate the stochastic effects of the wireless channel, a minimum mean-square error (MMSE) channel equalizer is integrated, which improves the detection accuracy by reducing the attack detection error by 21.6%. The RF transmission packet structure is modeled based on the IEEE 802.15.4 protocol. We conducted extensive over-the-air measurements under various attack modalities and scenarios using universal software radio peripheral (USRP) B210 software defined radios (SDRs) to validate the detection accuracy of our proposed hardware-based detection method for energy-constrained IoT systems at 915 MHz. Measurement results demonstrate an accuracy of 96.4% for detecting deceptive jammer and periodic jammer, closely aligning with the simulation results.