Traffic-Aware Agentic AI Framework for Semantic Validation of V2X Safety Messages

V2X communication is used to exchange safety information in real time manner in intelligent transport systems, but current methods of securing the V2X communication network are only concerned with ensuring the legitimacy of the message, without validating whether the alert itself is valid from a behavioral point of view. In order to resolve this problem, we proposed Score–Driven Adaptive Threshold Optimization for V2X Alert Validation approach to detect fake V2X alerts based on multi-vehicle behavior. An adaptive threshold calculation is conducted for evaluating the semantic consistency score with respect to varying traffic conditions and evidence. To mitigate the risk from any possible attack by a node, a trust-based evidential integration technique is included in the process for accurate decision making. An agentic feedback cycle is also implemented in this framework, allowing for dynamic adaptation of the threshold and the trust scores. Experimental analysis in low, moderate, and heavy traffic conditions shows that the framework achieves high detection accuracy, increased recall, F1-score balance, low false positives, and resilience to replay, spoofing, and Sybil attacks.