Intelligent Security Monitoring: Machine Learning-based Intrusion Detection
M. Rekha, P. Shobha Rani, S Aashida, S Elakkiya, Ezhil Arasi S, A Amirtha · 2025
The rapid growth in the field of Information Technology has definitely caused a steep rise in the rate and sophistication of cyber-attacks; therefore, cybersecurity is an issue of major concern. Recently, Machine Learning and Deep Learning methodologies have turned out to be really potent tools for network intrusion detection. This paper presents a detailed review of various ML and DL models used in Intrusion Detection Systems (IDS). The models like Naïve Bayes, Artificial Neural Networks, Support Vector Machines, Decision Tree, K-Nearest Neighbors, and methods based on Reinforcement Learning are covered. In this paper, the MRF model and examine the model with other ML and DL models are focused. In this analysis, the authors use the KDD dataset, a very popular standard in intrusion detection research. This dataset has different attributes divided into four main classes. Such as Content, Host, Basic, and Traffic. All these are vital classes in deciding the false alarm rate and detection rate of IDS. The empirical evidence that the MRF model outperforms traditional models in both measures of accuracy and reliability are provided. How contribution analysis for each class of attributes towards DR and FAR can be utilized in optimizing the MRF model for better performance of intrusion detection systems are shown. The results obtained in this research work have been pointing toward the importance of choosing suitable data attributes and models so that maximum DR can be achieved with a minimum FAR, which would increase the effectiveness of IDS. The proposed Modified Random Forest-based Intrusion Detection System uses advanced feature selection and preprocessing to improve detection accuracy and reduce false alarms. The real-time adaptability is done by dynamic selection of parameters in order to prevent evolving cyber threats.