Adversarial Machine Learning in Wireless Communications Using RF Data: A Review

Abstract

Machine learning (ML) provides effective means to learn from spectrum data and solve complex tasks involved in wireless communications. Supported by recent advances in computational resources and algorithmic designs, deep learning (DL) has found success in performing various wireless communication tasks such as signal recognition, spectrum sensing and waveform design. However, ML in general and DL in particular have been found vulnerable to manipulations thus giving rise to a field of study called adversarial machine learning (AML). Although AML has been extensively studied in other data domains such as computer vision and natural language processing, research for AML in the wireless communications domain is still in its early stage. This paper presents a comprehensive review of the latest research efforts focused on AML in wireless communications while accounting for the unique characteristics of wireless systems. First, the background of AML attacks on deep neural networks is discussed and a taxonomy of AML attack types is provided. Various methods of generating adversarial examples and attack mechanisms are also described. In addition, an holistic survey of existing research on AML attacks for various wireless communication problems as well as the corresponding defense mechanisms in the wireless domain are presented. Finally, as new attacks and defense techniques are developed, recent research trends and the overarching future outlook for AML in nextgeneration wireless communications are discussed.

Security Enhancement of mmWave MIMO Wireless Communication System Using Adversarial Training

“…It entails putting the model through training on both the actual training data and adversarial instances, which are intentionally misleading inputs. When the model is trained with these adversarial examples, it becomes more resilient and is better able to generalize its predictions to deal with comparable attacks in real‐world situations [15]. In this research work, adversarial training is done for untargeted, white‐box, gradient‐based adversarial attacks.…”

How to Attack and Defend NextG Radio Access Network Slicing With Reinforcement Learning

“…To maximize the impact of jamming the RBs, we pursue an adversarial machine learning approach. Different types of attacks built upon adversarial machine learning have been studied in wireless communications [21], [22] such as exploratory (inference) attacks [23], [24], evasion (adversarial) attacks [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39] and their extensions to secure and covert communications against eavesdroppers [40], [41], [42], causative (poisoning) attacks [43], [44], [45], membership inference attacks [46], [47], Trojan attacks [48], and spoofing attacks [49], [50], [51] that have been launched against various spectrum sensors and wireless signal (such as modulation) classifiers. Adversarial machine learning has also been considered for NextG by studying evasion and spoofing attacks on deep neural networks (without reinforcement learning) used for NextG spectrum sharing and NextG signal authentication [52].…”

Ameaças e Vulnerabilidades em Open RAN: Desafios e Soluções

“…Entretanto, um atacante que observa o meio e envia um sinal malicioso gera perturbações na entrada de modelos em treinamento ou em produção para efetuar ataques. Os ataques e defesas utilizados em outros problemas de aprendizado de máquina não são necessariamente aplicáveis em comunicações sem fio, devido à natureza dinâmica do canal de comunicação [Adesina et al, 2023]. Os trabalhos relacionados a esse tópico são muito heterogêneos quanto ao tipo de tarefa de aprendizado, aos tipos de ataque, aos mecanismos de defesa e às tecnologias de radiofrequência (RF) utilizadas.…”