Human-Centered Explainable AI for Security Enhancement: A Deep Intrusion Detection Framework

Explainable AI makes cybersecurity systems more trustworthy: The paper demonstrates that adding interpretability tools (specifically SHAP) to AI-powered intrusion detection systems allows security analysts to understand why the system flagged a threat, rather than treating it as a "black box." This transparency is essential for professionals who need to validate and act on security alerts.

High accuracy alone isn't enough for responsible deployment: While the deep learning models achieved 99% accuracy, the researchers recognized this wasn't sufficient for real-world use. They conducted a trust-focused expert survey with security professionals to evaluate whether the system was actually reliable and usable in practice—highlighting that responsible AI requires human validation beyond performance metrics.

Specific features matter for security decisions: The SHAP analysis identified which network traffic patterns were most influential in detecting intrusions (such as server error rates and destination host error rates). Making these decision factors explicit helps security teams understand what the system is actually monitoring and whether those factors align with genuine threats.

Human-centered design is critical for AI in security: The paper's emphasis on combining technical performance with expert surveys and interactive user interfaces reflects that cybersecurity AI systems must be designed with the people who use them in mind—not just optimized for algorithmic performance.