Towards Robust Network Security: Evaluating Machine Learning Algorithms for Intrusion Detection
Hafiz Burhan Ul Haq, Rabia Younis, Muhammad Shujat Ali · Decision Making Advances · 2025
The constant growth of cyber threats has made network intrusion detection systems (NIDS) more crucial. Targeting anomalous behavior in a network is challenging because of the large number of features that exist. Consequently, the accuracy is affected, and there will be a greater chance of less reliability in the network. This study overcomes the limitations of traditional NIDS by using multiple machine learning (ML) algorithms to enhance intrusion detection capabilities. The efficacy of many ML algorithms in the context of NIDS is examined by paying special attention to their ability to detect intrusion based on features. Similarly, experimental analysis is conducted using a publicly available large dataset containing 41 features, whereby six algorithms were compared: AdaBoost, Gaussian Naive Bayes, Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbors, and Multinomial Naive Bayes. Resultantly, SVM achieves the lowest accuracy at 53.15%. RF performs the best with a 99.78% accuracy rate. In addition, a comparative analysis is also performed, which is crucial for practitioners in the industry who want to implement effective NIDS.