Rabbit and Tortoise Optimization Algorithm with Mutual Information Based Adaptive Strategy for Network Intrusion Detection

In the modern era of highly interconnectedness, data and information are constantly transmitted over networks. Ensuring the security of confidential information and protecting computer systems from network threats has become very important. Therefore, it is important to develop an effective network intrusion detection system (NIDS) using optimal features. These optimal features can be identified through computational intelligence by learning patterns and relationships among features using machine learning techniques. This paper presents a rabbit and tortoise optimization technique for selecting optimal features. For evaluation, the UNSW-NB15 dataset is utilized. The optimization results achieve an accuracy of 94.12% for binary classification and 93.92% for multiclass classification, with 26 optimal features selected from the entire feature set. To improve the approach, an adaptive strategy based on mutual information is used to control the number of optimal features. This strategy, together with the Rabbit and Tortoise algorithm, improves the accuracy, showing 94.69% for binary classification and 94.03% for multiclass classification, while reducing the number of selected features to 9 only. The comparative performance analysis shows that the proposed feature selection method outperforms other state-of-the-art methods, providing more accurate and reliable results in identifying cyber threats. In addition, the relationship plot between the number of optimal features and the accuracy of the model shows that selecting only 9 features is effective in achieving high accuracy in detecting and predicting cyber attacks.