Autoencoder-AHP-Driven Learning for Blackhole Attack Detection in UAV Networks

The security of Flying Ad-Hoc Networks (FANETs) is compromised by blackhole attacks, where malicious UAVs drop all data packets by falsely advertising optimal routes. This study proposes Autoencoder-AHP-Driven Learning for Blackhole Attack Detection in UAV Networks, an intrusion detection system that integrates an Autoencoder-driven Analytic Hierarchy Process (AHP) with Random Forest (RF) to detect and isolate blackhole nodes.The Autoencoder is utilized to determine the importance of each security metric rather than for dimensionality reduction. It assigns importance scores to key features: Packet Drop Ratio (PDR), Sequence Number Anomaly (SNA), RREQ Reply Ratio (RREP-RR), and Hop Count Variation (HCV), which are essential for identifying abnormal packet loss, route inconsistencies, excessive fake replies, and sudden topology changes. The feature importance scores obtained from the Autoencoder are then processed by AHP to compute optimal feature weights, ensuring an objective prioritization of the most relevant metrics. The AHP-weighted dataset is subsequently classified using Random Forest, achieving high detection accuracy with minimal computational overhead.Experimental results demonstrate that the proposed AHP-Driven Machine Learning approach achieves 99.2% accuracy, making it well-suited for real-time UAV security. This approach enhances detection precision while maintaining efficiency, providing a scalable solution to mitigate blackhole attacks in FANETs.