Addressing the dual challenges of platoon safety and communication resource optimization for connected vehicle platoons (CVPs) subject to False Data Injection (FDI) attacks in highway scenarios, this paper proposes a resilient control strategy that integrates an adaptive neural network with a dynamic event-triggered mechanism. The proposed strategy incorporates explicit modeling of attack parameters and platoon error constraints to unify attack mitigation and communication optimization. Specifically, the main contributions are three-fold: 1) A distributed output feedback framework incorporating time-varying FDI attack characteristics is established, alongside a Radial Basis Function (RBF) neural network-based state observer. By explicitly modeling attack parameters within the observer structure and the adaptive update laws, unmeasurable vehicle states are reconstructed via the neural network"s dynamic approximation of the attack signals. 2) A dynamic event-triggered condition, driven by both attack intensity and neural approximation errors, is developed. By introducing a dynamic threshold variable regulated by attack intensity negative feedback, a closed-loop adaptive mechanism—characterized by "attack enhancement → dynamic variable decay → trigger threshold reduction → communication frequency increase"—is formulated, achieving a dynamic balance between attack mitigation and communication efficiency. 3) An attack-parameter-dependent Lyapunov functional is constructed, accounting for neural weight errors, state estimation errors, and dynamic triggering variables. Utilizing Linear Matrix Inequalities (LMIs), the semi-global uniform ultimate boundedness (SGUUB) of all signals in the closed-loop system is rigorously proven, effectively reducing the conservatism of the stability criteria. Theoretical analysis confirms that the proposed strategy ensures platoon safety under both attack and communication constraints. Simulation results demonstrate that under time-varying FDI attacks, the maximum platoon error remains within 2 m, and the communication load is reduced by 96.4% compared to traditional time-triggered mechanisms. Furthermore, both the convergence speed and control accuracy surpass traditional decoupled design methods, validating the strategy’s effectiveness and practical engineering potential.