<p>Computation offloading in Internet of Vehicles (IoV) provides the powerful technical support for the new on-board services such as automatic driving and unmanned driving. However, the poor network condition in some areas led to delayed data transmission, and bring considerable challenges for the implementation of edge computing in vehicular networks. Moreover, vehicular networks are highly vulnerable to cyberattacks, such as data theft, information tampering, and unauthorized system access, which pose severe security risks. UAV as an auxiliary component of IoV systems offers significant advantages such as its high flexibility, low cost, ease of deployment, and so on. Blockchain can prevent the tampering or forging of information. So in this paper, we propose a secure and collaborative computation offloading framework with dynamic service caching in blockchain-empowered UAV-assisted vehicular networks. This framework consists of offloading decision making, dynamic service caching and caching updating decisions, computation resource allocation, and blockchain consensus mechanism. Under the constraint of UAV’s energy consumption and limited storage space, UAV’s caching management and updates are performed, while the Practical Byzantine Fault Tolerance (PBFT) protocol on Mobile Edge Computing (MEC) is employed to achieve low consensus latency and record valid offloading information. Then an NP-hard problem is formulated with the aim to minimize the total system delay and is solved by the Soft Actor Critic (SAC) reinforcement learning algorithm. Finally, several simulation experiments were conducted to confirm the effectiveness of the suggested approach.</p>