Exploring Explainable AI, Security and Beyond : A Comprehensive Review

Abstract

The paper dives into the transformation of security with the integration of machine learning (ML) and the associated challenges. It highlights how AI’s incorporation in network security brings both promise and complexities, emphasizing the need to align perceived benefits with actual capabilities. Explainable AI (XAI) emerges as a crucial tool, offering transparency despite facing ongoing challenges and necessitating continual advancements. The pursuit of Explainable AI (XAI) and tools like AI Explainability 360 demonstrate strengths but grapple with understanding and methodological gaps. Specific techniques such as LIME, GNNEXPLAINER, and object recognition in Deep Reinforcement Learning show promise but encounter challenges like scalability and adaptability. Across these domains, understanding the multifaceted landscape becomes pivotal for leveraging’s potential while addressing critical challenges in security and explainability.