Demystifying the “Black Box” of Deep Learning in Cybersecurity: Integrating Explainable AI (XAI) to Augment the Reliability of Intrusion Detection Systems
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
The rapid and continuous rise of sophisticated cyberattacks has driven the adoption of Deep Learning models in Intrusion Detection Systems (IDS). While these models deliver superior detection performance compared to traditional approaches, their “black-box” nature poses a significant barrier to real-world deployment. Network administrators often struggle to trust system-generated alerts when they cannot understand the reasoning behind them. This paper proposes a comprehensive solution to address this challenge by integrating Explainable Artificial Intelligence (XAI) techniques into IDS. Specifically, we apply SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-Agnostic Explanations) to Deep Neural Network (DNN) models trained on benchmark datasets, namely NSL-KDD and UNSW-NB15. The results demonstrate not only the model’s high effectiveness in detecting various types of cyberattacks but also its ability to provide detailed explanations at both global and local levels. These insights enable network administrators to better analyze the root causes of alerts, thereby improving the reliability and transparency of cybersecurity defense systems.
Fields of Study & Subjects
Network Security and Intrusion Detection
Explainable Artificial Intelligence (XAI)
Adversarial Robustness in Machine Learning
Publication details
- Journal / source
- East African Scholars Journal of Engineering and Computer Sciences
- Volume
- 9
- Issue
- 02
- Pages
- 15–19
- Publication date
- 30 March 2026
- Language
- EN