Intrusion Detection System Analysis to Improve Computer Network Security
Barlyan Kurdianto, Yohanzah Febriyanto, Yustian Servanda · Journal of Artificial Intelligence and Engineering Applications (JAIEA) · 2025
This study analyzes intrusion detection systems (IDS) as a vital component in the security of modern computer networks that face increasingly complex cyber threats. Through the Systematic Literature Review approach of 478 publications during 2019-2024, it was found that the CNN-LSTM hybrid model achieved a detection accuracy of 97.3% on the NSL-KDD dataset, far surpassing conventional signature-based methods. The implementation of anomaly-based IDS on Indonesian government infrastructure has identified 45% of attacks that are not detected by traditional solutions, with a reduction in incident response time from 24 hours to 3.5 hours. Federated learning technology for heterogeneous IoT environments increases detection accuracy by 18.7% while reducing network load by up to 76%, while integration with blockchain reduces incident investigation time by 67%. Explainable AI-based frameworks increase security team confidence by 43% and reduce alert fatigue by 38%. The reinforcement learning-based IDS system showed autonomous adaptability with an increase in F1-score from 0.87 to 0.96 without manual intervention. The cost-benefit analysis shows a positive return on Security Investment with an average breakeven point achieved in 14-19 months. This research provides the foundation for the development of an adaptive, contextual, and integrated intrusion detection system to deal with the evolution of contemporary cyber threats. Keywords: Intrusion detection systems, network security, artificial intellige