Explainable Artificial Intelligence (XAI) for Detecting Cybersecurity Threats: Enhancing Trust, Transparency, and Accuracy in AI-Driven Security Systems

Artificial intelligence (AI) is increasingly used in cybersecurity to identify unusual patterns, detect complex attacks, and support responses to emerging threats. However, the limited transparency of many AI-powered security systems raises concerns about trust, accountability, and reliability in security operations. Explainable artificial intelligence (XAI) seeks to make AI-generated predictions and decisions more understandable to human users. Its usefulness depends on the fidelity, clarity, and security of the explanations it provides. This paper examines how XAI may support trust and transparency in AI-driven cybersecurity threat detection and considers its relationship with predictive accuracy and human decision-making. The study adopts a qualitative, interpretivist, exploratory, and documentary approach based on secondary evidence, including previous research, cybersecurity standards, technical publications, and relevant institutional reports. Thematic analysis is used to identify recurring patterns, conceptual relationships, implementation challenges, and governance implications. Five interconnected dimensions are examined: (1) the interpretability of threat-detection decisions; (2) informed trust and human oversight; (3) false-positive and false-negative detection; (4) complex and evolving cyber threats; and (5) governance concerns involving explanation fidelity, privacy, adversarial manipulation, and model complexity. A central finding is that explainability does not necessarily improve predictive accuracy or cybersecurity performance. Rather, it can give security analysts greater insight into system outputs, helping them assess automated recommendations and make informed decisions within appropriate operational and governance limits. The study offers a conceptual framework for understanding how humans and AI can work together in cybersecurity threat detection and how AI can be governed and used responsibly. It concludes that explainability should be combined with comprehensive security assessments, privacy safeguards, and meaningful human oversight to support more transparent, trustworthy, and reliable cybersecurity systems.