Explainable AI and Statistical Analysis for Enhanced Reliability in Unmanned Aerial System Intrusion Detection

The research paper explores critical vulnerabilities in AI-based intrusion detection systems specifically designed for Unmanned Aerial Systems. By integrating Explainable AI with rigorous statistical analysis, the authors aim to solve the black-box nature of traditional detection models that often fail to provide actionable security insights. This approach ensures that anomalies are not only detected but also interpretable for security operators to facilitate faster response times. The proposed methodology identifies specific areas where existing detection systems are susceptible to adversarial manipulation or environmental noise. By updating the scope of vulnerability assessment, the study highlights how previous security frameworks may lack the granularity required for modern aerial operations. Implementing these findings requires a shift in how engineers validate AI model outputs during real-time flight monitoring to ensure operational safety. The technical implementation focuses on bridging the gap between raw statistical data and human-understandable explanations. This ensures that security patches and system updates are targeted effectively rather than applied as broad, inefficient fixes. Engineers working with autonomous systems should review the provided statistical models to assess whether their current intrusion detection pipelines require architectural adjustments or enhanced logging capabilities.