Designing Learning-Based Adversarial Attacks to (MIMO-)OFDM Systems With Adaptive Modulation

Abstract

With the development of 5G, 6G, and beyond, wireless communications have taken a vital role in our daily life. The wide adoption of wireless devices and applications makes security of the wireless transmissions more critical. Many attacks aiming at wireless communications have been proposed (e.g., jamming attack). Orthogonal Frequency Division Multiplexing (OFDM) together with multi-input and multi-output (MIMO) design and adaptive modulation (AM) becomes an important technique to achieve high data rates. However, most of the attacks are not tailored for OFDM based systems. In this paper, we reveal a potential security issue of (MIMO-)OFDM-AM systems by developing three learning-based attack methods to (MIMO-)OFDM-AM systems. Based on the principle of AM, channel state information (CSI) is inferred from the modulation type of each subcarrier without any prior knowledge of the desired transmission. With the inferred CSI, we develop three adversarial attacking methods by maximizing error rate, minimizing capacity, or maximizing outage probability. Simulation results show that our learning-based method can detect the modulation type of each subcarrier in high accuracy and then successfully infer the CSI range. Furthermore, simulations also demonstrate that the proposed attacks cause severe performance degradation of (MIMO-)OFDM-AM systems (e.g., error rate, capacity) with fairly low power.