<p>The large-scale deployment of mobile edge computing (MEC) has enabled efficient task offloading in vehicular networks. However, it also introduces spatiotemporal privacy leakage problems, particularly concerning location and trajectory data. To address these challenges, we propose a joint optimization model of trajectory obfuscation with differential privacy and task offloading in mobile edge computing (JOM-TODPTO). Firstly, a trajectory position prediction mechanism based on spatiotemporal perception was proposed. In this mechanism, the trajectory data is described as a geographic adjacency matrix, the spatiotemporal convolutional network is used to extract the trajectory data features, and the subsequent location is predicted according to the historical trajectory. Secondly, a personalized privacy protection mechanism was proposed. The sensitivity of the location point is evaluated according to its predictability and importance. Then, the W-window mechanism is used to assign personalized privacy to the location point. Then, the differential privacy technology is used to perturb the real location to the perturbed area. Finally, a task offloading strategy generation mechanism based on a genetic algorithm was proposed. This model relies on high-performance computing capabilities to process large-scale spatiotemporal trajectory data and achieves real-time task offloading through parallel distributed optimization, meeting the millisecond-level response requirements of scenarios such as autonomous driving. Simulation results show that, compared with other schemes, JOM-TODPTO improves privacy protection performance while ensuring data availability and keeping the average computation cost low.</p>