Deep Cross-Check Q-Learning for Jamming Mitigation in Wireless Networks

Due to their open and shared nature, wireless networks are susceptible to jamming attacks. Collaborative efforts among legitimate users are explored to enhance network resilience and security. Distributed learning algorithms and deep reinforcement learning techniques foster effective cooperation and resource optimization, bolstering network performance. We introduce “Deep Cross Check Q-learning,” a modified distributed multi-agent reinforcement learning algorithm in response to persistent wireless network attacks. Legitimate users leverage observations of peers’ behavior to make informed decisions, factoring in their Deep Q-Network and peer actions. Our method demonstrates significant efficacy against both reactive and Deep Q-learning-based jamming attacks, resulting in improved signal-to-noise ratios and reduced mutual interference.

I. Introduction

Wireless networks’ shared and open nature makes them susceptible to various attacks, which can undermine their security and reliability. Attackers can intercept, eavesdrop, or disrupt communication, resulting in severe privacy breaches or service outages. Among these attacks, jamming attacks are especially pernicious, as they aim to overload the wireless network with malicious signals to disrupt legitimate communication links, causing users to experience difficulties transmitting or receiving data.