Assessing Machine Learning Effectiveness in Intrusion Detection Systems

The increased complexity of cyber attacks and advanced persistent threat (APT) scenarios has elevated the importance of Intrusion Detection Systems (IDS) in securing networks from maliciousness. Because of the fast-changing nature of cyber threats and attacks, most traditional techniques, including rules-based systems and signatures, are no longer providing adequate protection against sophisticated cyber-attacks. This study examines the use of various machine learning algorithms (decision tree [DT], k-nearest neighbor [KNN], support vector machine [SVM], and logistic regression [LR]) using the NSL-KDD dataset for intrusion detection classification. Overall, the proposed approach produces results indicating that Machine Learning Algorithms outperform traditional methods, exhibit high classification accuracy, and demonstrate capability of classifying known and unknown attacks. As a result, this research enhances the performance of intrusion detection systems (IDS) by ensuring an effective and versatile solution to securing networks. Performance evaluations of several machine learning algorithms showed a high level of success (99.58%, 98.33%, 97.70%, and 91.96%). Among the four evaluated models, the LR model had the lowest level of performance accuracy while the DT model had the highest level of accuracy.