Vehicular Ad Hoc Networks (VANETs) are vulnerable to intelligent attackers who exploit the homogeneous treatment of traffic events in existing trust models. These attackers accumulate reputation by reporting correctly on low-priority events and then inject false data during safety-critical situations - a strategy that current approaches cannot detect because they ignore event severity and location criticality in trust calculations. This paper addresses this gap through three contributions. First, it introduces event-aware and location-aware intelligent attack models, which have not been formally defined or simulated in prior work. Second, it proposes an asymmetric local trust mechanism where penalties scale with event and location severity while rewards follow an asymptotic model, making trust difficult to regain after misuse. Third, it adapts Dempster-Shafer Theory for global trust fusion using Yager's combination rule - assigning conflicting evidence to uncertainty rather than forcing premature decisions - combined with sequential source-reliability ordering and an asymmetric risk accentuation mechanism. Simulations using OMNeT++, Veins, and SUMO compare the proposed system (IPEK) against MDT and TCEMD under attacker densities of 15-35 percent. IPEK maintained 0 percent False Positive Rate across all scenarios, meaning no honest vehicle was wrongly revoked, while sustaining Recall above 75 percent and F1-scores exceeding 0.86. These results demonstrate that integrating context-awareness into both attack modeling and trust evaluation significantly outperforms symmetric approaches against strategic adversaries.

翻译：车载自组织网络（VANETs）易受智能攻击者威胁，此类攻击者利用现有信任模型对交通事件的同质化处理方式：通过正确报告低优先级事件积累信誉，随后在安全关键场景中注入虚假数据——当前方法因忽略事件严重程度和位置关键性而无法检测这种策略。本文通过三项贡献填补这一空白。首先，提出事件感知与位置感知的智能攻击模型，这在以往工作中未被正式定义或仿真。其次，设计非对称局部信任机制，其中惩罚力度随事件与位置严重性递增，而奖励遵循渐近模型，使信任在滥用后难以恢复。第三，采用Yager合成规则改进Dempster-Shafer理论以融合全局信任——将冲突证据分配至不确定性而非强制决策——并结合顺序源可靠性排序与非对称风险强化机制。基于OMNeT++、Veins和SUM0的仿真在15-35%攻击者密度下将所提系统（IPEK）与MDT、TCEMD进行对比。IPEK在所有场景中保持0%的假阳性率（即无诚实的车辆被错误撤销），同时召回率维持在75%以上，F1分数超过0.86。结果表明，将上下文感知能力融入攻击建模与信任评估，在面对策略性对手时显著优于对称方法。