2020/11/03 by Samurdhi Karunaratne, Karunaratne, Samurdhi, Enes Krijestorac +3 · 2 citations
Computer Science · Engineering · #Cryptography and Security (cs.CR) #FOS: Computer and information sciences #FOS: Electrical engineering #Full-Duplex Wireless Communications #Hate Speech and Cyberbullying Detection #Signal Processing (eess.SP) #Wireless Signal Modulation Classification #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2011.01538
openalex publication_date 2020/11/03 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Physical layer authentication relies on detecting unique imperfections in\nsignals transmitted by radio devices to isolate their fingerprint. Recently,\ndeep learning-based authenticators have increasingly been proposed to classify\ndevices using these fingerprints, as they achieve higher accuracies compared to\ntraditional approaches. However, it has been shown in other domains that adding\ncarefully crafted perturbations to legitimate inputs can fool such classifiers.\nThis can undermine the security provided by the authenticator. Unlike\nadversarial attacks applied in other domains, an adversary has no control over\nthe propagation environment. Therefore, to investigate the severity of this\ntype of attack in wireless communications, we consider an unauthorized\ntransmitter attempting to have its signals classified as authorized by a deep\nlearning-based authenticator. We demonstrate a reinforcement learning-based\nattack where the impersonator--using only the authenticator's binary\nauthentication decision--distorts its signals in order to penetrate the system.\nExtensive simulations and experiments on a software-defined radio testbed\nindicate that at appropriate channel conditions and bounded by a maximum\ndistortion level, it is possible to fool the authenticator reliably at more\nthan 90% success rate.\n