Real-time anomaly detection in smart vehicle-to-UAV networks for disaster management

Abstract

In disaster situations, conventional vehicular communication networks often face heavy congestion, which hinders the effectiveness of Vehicle-to-Vehicle (V2V) communication. To overcome this issue, Vehicle-to-Unmanned Aerial Vehicle (V2U) communication is a crucial alternative, offering an expanded network infrastructure for real-time information sharing. Nonetheless, both V2V and V2U networks are vulnerable to cyber-physical disruptions caused by malicious attacks, signal interference, and environmental factors. This paper introduces an advanced anomaly detection framework tailored for disaster-response vehicular networks, which combines a discrete-time Markov chain (DTMC) with machine learning (ML) methods. The model employs DTMC to define normal transmission behavior while adaptively modifying state transition probabilities through ML techniques using real-time data. The simulations in MATLAB validate the proposed method by analyzing log-likelihood maneuver patterns and evaluating detection performance with Receiver Operating Characteristic (ROC) curves. Our findings reveal that the hybrid DTMC-ML model successfully detects anomalies in both V2V and V2U networks, achieving a high true positive rate while reducing false alarms. This research aids in advancing resilient vehicular communication systems for disaster response, thereby improving the reliability and security of intelligent transportation networks in extreme situations.