A Practical and Explainable Network Anomaly Detection System for Enhanced Cybersecurity

Network Intrusion Detection Systems (NIDS) are pivotal in safeguarding digital infrastructures against evolving cyber threats. This work presents a novel explainable anomaly detection framework leveraging a 1D Convolutional Neural Network (1D-CNN) for anomaly classification. The proposed model attains a training accuracy of 99.42% and a validation accuracy of 99.16% on the benchmark KDDCUP’99 dataset. To address the critical challenge of interpretability, we employ SHapley Additive exPlanations (SHAP) plots, SHAP force plots, and heatmap visualizations to better understand the decision-making process of the model. By focusing on sequential data patterns, the 1D-CNN effectively classifies network traffic, while the integration of explainability techniques ensures comprehensibility and trustworthiness. This framework bridges the gap between high-performance anomaly detection and model interpretability, offering a robust, deployable solution for real-world network security applications.