An Unmanned Network Intrusion Detection Model Based on Deep Reinforcement Learning
Kezhou Ren, Maohuan Wang, Yifan Zeng, Yingchao Zhang · 2022 IEEE International Conference on Unmanned Systems (ICUS) · 2022
Network assaults pose significant security concerns to network services. Hence it is imperative to use new technical ways to enhance the performance of intrusion detection systems. Several reinforcement learning algorithms for network intrusion systems (e.g., Markov and other methods) have been introduced in recent years to satisfy the unmanned and intelligent requirements of IDS. This paper presents a deep Q-learning-based network intrusion detection model that provides continuous automatic learning capability for network environments based on reinforcement learning and incorporates a deep feed-forward neural network approach that can detect various types of network attacks. Experiments were done using CSE-CIC-IDS2018 as a dataset to test the model's performance, which provides a complete set of real network traffic. Experimental results demonstrate that the proposed model successfully detects network threats, outperforms conventional machine learning techniques, and has promising possibilities for unmanned IDS in complex network settings.