Explainable AI (XAI) for Cybersecurity Decision-Making Using SHAP and LIME for Transparent Threat Detection
Shantha Visalakshi Upendran, S. Karthiyayini, Dinesh V. Jamthe · 2025
The increasing complexity and sophistication of cyber threats have necessitated the integration of Explainable Artificial Intelligence (XAI) into cybersecurity frameworks to enhance transparency, trust, and decision-making. Traditional black-box machine learning models, despite their high accuracy, pose significant challenges in understanding threat detection mechanisms, leading to reduced interpretability and limited adoption in critical security applications. This book chapter explores the role of XAI techniques, specifically Shapley Additive Explanations (SHAP) and Local Interpretable Model-Agnostic Explanations (LIME), in improving the explainability of AI-driven cyber defense systems. A detailed analysis of computational efficiency, real-time applicability, and scalability challenges associated with SHAP and LIME in large-scale cybersecurity environments is provided., the chapter introduces hardware-accelerated approaches, such as FPGA-based optimization, to mitigate computational overhead while ensuring rapid and interpretable threat detection. reinforcement learning-based optimization for explainability is examined to enhance adaptive security mechanisms in dynamic threat landscapes. The integration of XAI-driven security information and event management (SIEM) systems is also discussed to bridge the gap between automated cyber threat detection and human-centric decision-making. This chapter provides a comprehensive exploration of state-of-the-art methodologies, challenges, and future research directions in the domain of XAI for cybersecurity, with a focus on balancing detection accuracy, computational efficiency, and interpretability.