Year: 2026
Type: article
Abstract: Intrusion Detection Systems (IDS) are critical for protecting computer networks from cyber threats. Traditional signature-based detection systems are inadequate against novel and evolving attacks. This paper proposes an Explainable AI-based IDS using a Long Short-Term Memory (LSTM) neural network trained on the CIC-IDS2017 dataset. The model classifies network traffic as normal or malicious with high accuracy. To enhance transparency, SHAP and LIME explainability techniques are integrated, providing both global feature importance and local prediction explanations. The system is deployed using Django, where users input network features and receive real-time predictions with confidence scores and visual explanation graphs. Experimental results demonstrate 96.8% detection accuracy with an AUC of 0.98, while SHAP analysis reveals that flow duration, packet length, and flag counts are the most influential features for intrusion detection.
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7216955118
