Authors:
Christoph Wech, K. Srinarayani, Chiara Brüggemann, Andrin Bertschi, Fabian Richter, Deepa Nair
Addresses:
Department of Computer Science, Secure and Trustworthy Systems Group, ETH Zurich, Zürich, Switzerland. Department of Computer Science in Artificial Intelligence and Machine Learning, SRM Institute of Science and Technology, Ramapuram, Chennai, Tamil Nadu, India. Department of Physics, Palffy-Buß Group, University of Wurzburg, Würzburg, Bavaria, Germany. Department of Logistics, Tampines Industries Pte. Ltd., East Region, Singapore.
Cybersecurity threats continue to evolve in complexity, making accurate and interpretable intrusion detection systems essential for modern network environments. This paper presents an explainable cyber attack detection framework utilizing machine learning techniques and SHapley Additive exPlanations (SHAP) for transparent decision-making. The proposed framework was evaluated on the CIC-IDS2017 dataset, which contains both benign and malicious network traffic representing various cyber attack scenarios. Three machine learning approaches were investigated, namely Isolation Forest, One-Class Support Vector Machine (One-Class SVM), and Random Forest. Isolation Forest and One-Class SVM were employed as unsupervised anomaly detection techniques, while Random Forest was used as a supervised attack classification model. Experimental results demonstrated that Random Forest significantly outperformed the anomaly detection approaches, achieving a Precision of 99.64%, Recall of 98.86%, F1-Score of 99.25%, and ROC-AUC of 99.98%. To improve transparency, SHAP was integrated to identify and explain the most influential network traffic features contributing to attack detection. The analysis revealed that destination port characteristics and packet length statistics were among the most significant indicators of malicious activity. The proposed framework combines high detection performance with explainability, making it suitable for practical cybersecurity applications and security monitoring environments.
Keywords: Intrusion Detection System; Cyber Attack Detection; Machine Learning (ML); Random Forest; Isolation Forest; Anomaly Detection; Shapley Additive Explanations (SHAP).
Received on: 30/08/2025, Revised on: 09/11/2025, Accepted on: 18/12/2025, Published on: 11/09/2026
DOI: 10.69888/FTSIS.2026.000804
FMDB Transactions on Sustainable Intelligence and Security, 2026 Vol. 1 No. 3, Pages: 168-181