Interpretable Anomaly-Based DDoS Detection in AI-RAN with XAI and LLMs

AI security systems can now explain their decisions in plain language: The paper demonstrates that Large Language Models can translate complex technical explanations from AI security tools into natural language that non-experts can understand, making it possible for more people to comprehend why a system flagged a potential attack.

High accuracy is achievable for detecting network attacks: The proposed system achieved an F1-score greater than 0.96 in detecting DDoS attacks on 5G networks, suggesting that AI-based security tools can reliably identify malicious activity when properly designed and trained on real network data.

Explainability methods are essential for trustworthy AI security: The paper emphasizes that using explainability techniques (LIME and SHAP) alongside AI models is critical for security applications, as it allows human operators to verify that the system is making decisions for the right reasons rather than relying on opaque "black box" predictions.

Future network infrastructure requires interpretable AI safeguards: As 5G and 6G networks become more automated and intelligent, the paper highlights that security systems protecting these networks must be both accurate and transparent—addressing a key gap where many current AI security tools lack adequate explainability.

Real-world testing validates responsible AI design: The framework was tested on actual 5G network data rather than simulations alone, demonstrating that responsible AI principles (interpretability + accuracy) can be implemented in production security systems, not just theoretical research.