An Improved Hybrid Intrusion Detection Approach

Abstract

In today’s modern era, our always-connected world is lined with smart devices which is a double-edged sword. Although it is tremendously convenient, on the other hand, our data and systems are progressively exposed to a growing army of hackers. In this regard, the need for a reliable IDS continues to grow. The reason is that intrusion detection systems is vital not only to protect data but to prevent computer systems from unauthorized access and cyber-threats. Conventional IDSs limitations lie in the dependence on the signature base detection, which renders them unable to identify unknown and harmful threats. In many ways, machine learning is a promising approach to the identification of such malicious activity through the development of efficient, high-performing solution. In this study, we propose a different hybrid approach that combines two of the most well-known methods of ML; the J48 and Random Forest. The proposed approach is also made more effective through the use of a recursive feature elimination process. This process selects the ideal subset of relevant features and increases the performance of the model. The proposed approach has been examined and tested using the NSL-KDD intrusion detection benchmark datasets, which covers all kinds of intrusions. The results demonstrate the ability of proposed method to effectively detect different types of intrusions and well compete with other state-of-the-art intrusion detection methodologies.