Intelligent Outlier Detection with Optimal Deep Reinforcement Learning Model for Intrusion Detection

Intrusion detection system (IDS) acts as an essential part to detect malicious activity in the cyber domain. Earlier works on IDS are mainly based on statistical, machine learning (ML), and deep learning (DL) approaches. Since deep reinforcement learning (DRL) becomes an emerging research area, it can be employed for intrusion detection and thereby accomplishes security in the present digital world. This paper focuses on the design of intelligent outlier detection with optimal deep reinforcement learning (IOD-ODRL) technique for intrusion detection. The proposed IOD-ODRL technique encompasses an Isolation Forest (iForest) based outlier detection approach to eliminate the existence of outliers in the test data. Besides, optimal Q-learning based DRL technique is employed for the detection and classification of intrusions. Moreover, the learning rate of the DRL technique is optimally chosen by the use of sandpiper optimization (SPO) algorithm. Furthermore, the design of outlier detection and SPO based learning rate selection results in improved detection performance. In order to investigate the superior performance of the IOD-ODRL technique, a wide range of simulations take place on benchmark datasets. The experiment results indicated that the betterment of the IOD-ODRL model interms of several metrics.