From Trust Scores to Driver Decisions: An Abstaining Evidential Decision Layer for Event Validation in VANETs

Trust management in vehicular ad hoc networks is highly active, yet the decision that a driver needs, whether a reported event is real, has received far less attention; the schemes that make it force every decision into two outcomes, even on thin or conflicting evidence. This paper presents CREST, a vehicle-side event-decision framework that can withhold a decision its evidence does not support. Each vehicle weights the reports it holds by the trust, proximity and freshness of their senders, fuses them with Yager's rule, and decides only when the fused margin reaches an abstention threshold. Abstention is a waiting state: a vehicle still unsure at the event is resolved by an explicit application policy, which makes the trade-off between missed hazards and unnecessary detours visible. CREST requires only a scalar trust per reporter and is built here on an evidential trust model whose consensus and revocation steps are updated to support the decision layer. A single report is shown to change the decision margin by at most its weight, which bounds the influence of an attacker coalition by its trust. In a pre-registered simulation study with fourteen attacks, including position spoofing and trust-gated camouflage, and with all unspecified parameters of all systems tuned by the same procedure, CREST reduces the decision error of two published trust-based dissemination systems from 0.162 and 0.106 to 0.051. With the best-guess policy it also lowers both missed and unnecessary reroutes and outperforms the forced binary rule on both error types in every cell.