Vehicular networks have gained considerable attention in recent years, driven by the growing demand for improved road traffic efficiency, autonomous driving capabilities, and onboard entertainment services. Vehicle-to-everything (V2X) communications have emerged as an important aspect of vehicular communications, allowing seamless communication between vehicles and V2X-enabled entities nearby. However, the high mobility of vehicles, scalability issues, different QoS requirements of V2X links, such as varying levels of reliability and stringent latency requirements of certain vehicular applications, and increasing security and privacy concerns of exchanging sensitive data present major obstacles to reliable communications. This work addresses these challenges by proposing novel solutions that enhance vehicular networks’ efficiency, performance, and security. First, a graph-based optimization method is introduced to allocate network resources and fulfill diverse QoS requirements in vehicle-to-infrastructure (V2I) and vehicle-to-vehicle (V2V) links. Secondly, to address latency concerns, the Latency-Aware Mode Coordination (LAMOC) algorithm is proposed to minimize end-to-end delays, thereby substantially improving communication reliability in dynamic vehicular scenarios. Additionally, the Blockchain-Enabled Vehicular Edge Computing (BEVEC) framework is proposed to build upon the need for secure and efficient service delivery. BEVEC integrates a dual-layer verification process using a permissioned blockchain to ensure data integrity and a Deep Reinforcement Learning (DRL) algorithm that optimizes a tailored utility function. This synergy achieves timely and energy-efficient service delivery while improving reliability. Moreover, to address the multifaceted challenges of dynamic vehicular networks—ranging from energy efficiency to privacy and scalability—a multi-layer Permissioned Distributed Ledgers (PDL)-based Cooperative Decentralized Vehicular Edge Computing (PDL-CoDeVEC) framework is introduced. A Multi-Agent Deep Reinforcement Learning (MADRL) algorithm, named Multi-Agent Cooperative Task Coordination (MACTAC), is proposed to optimize the trade-off between local and offloaded task processing that enables vehicles to independently make computation decisions while preserving real-time performance through utility maximization. Finally, a dynamic rolling-horizon resource allocation algorithm combined with a Vickrey–Clarke–Groves (VCG) pricing mechanism is proposed to address the dual challenges of efficient resource allocation and fair incentive schemes in IoV networks. Extensive simulations were conducted across diverse scenarios to evaluate the proposed solutions. The results demonstrate notable latency and energy consumption reductions, as well as improved throughput and task completion rates. These findings validate the efficacy of the proposed mechanisms in real-time vehicular environments, supporting their potential for large-scale deployment while maintaining robust security and privacy safeguards.