An Efficient Hybridization of K-Means and Genetic Algorithm Based on Support Vector Machine for Cyber Intrusion Detection System

Intrusion Detection System (IDS) is a challenging cyberspace security technology to safeguard against a malicious threat. Although many soft computing approaches have been utilized to increment the effectiveness of IDS, it is a significant challenge for present-day intrusion detection classification algorithms to give and achieve high performance. The first significant challenge is that lots of needless, dispensable, superfluous, and meaningless data in high-dimensional datasets affect the IDS classification process. Secondly, attack patterns are also dynamic, requiring efficient classification and cyber-attacks prediction. Thirdly, a single classifier cannot work well to detect any form of attack. Lastly, the accuracy, detection rate (DR), and false alarm rate (FAR) are still significant issues to contend with. Thus, we propose an efficient hybridization technique in this paper to address these significant challenges. This paper proposes supervised and unsupervised learning techniques for detecting both known and unknown attacks. In the first line of this research, k-means clustering was applied to the normalized data to classify the data into normal and attack classes to resolve the dynamic nature of the attack patterns. Then, wrapper feature selection with a genetic algorithm (GA) was employed to address the needless and redundant dataset. Lastly, the classification of the inputted data from GA predictors was performed with a support vector machine (SVM). The analysis of the computational time needed for training and testing for use in time-critical applications was also carried out. The experimental results revealed a promising high accuracy of 99% with low FAR. The appealing benefits of the proposed model are its robustness, low computational cost, and also its impressive success in generalization by reducing possible overfitting.