Deep Reinforcement Learning-Based Intrusion Detection System for Next-Generation Wireless Networks
A. Arun, CH Hussaian Basha, M. Jamuna Rani, V. Jamuna, P. Vijayakumar, K.J. Jegadish Kumar · 2025
Of late, this is due to fast evolution of next generation wireless networks, such as 5G, and unprecedented growth in number of connected devices and heterogeneous network architectures. This proliferation greatly increases the surface area for potential cyber threats and conventional Intrusion Detection Systems (IDS) are not sufficient because they are static and signature based. In this paper, we present a novel IDS based on Deep Reinforcement Learning used on dynamic and complex wireless environments. The system uses the adaptive learning capability of DRL to obtain optimal defense strategies by having continuous interaction with the network environment, and having the attack patterns evolve. The proposed model is based on a Dueling Deep Q Network (Dueling-DQN) architecture fortified with the use of prioritized experience replay for this purposes. Experimental evaluations on benchmark wireless traffic datasets show that our DRL-IDS achieves much better performance in detecting known as well as zero day attacks to a level very close to the lower bound set by the ideal detector while containing minimal false positives. In addition, the system is able to adapt to real time, scale up and robust, which makes it a perfect solution for securing the future wireless communication infrastructures (e.g. smart cities, vehicular networks and IoT driven ecosystem).