The 5G mobile communication technology has facilitated the proliferation of intelligent applications, including autonomous driving and virtual reality, which pose stringent requirements for low latency and high bandwidth. A significant subset of these services demands substantial computational resources under rigorous delay constraints, rendering the processing of large-scale tasks within limited time frames a considerable challenge. Furthermore, mobile devices are typically constrained by their limited battery capacity, and the elevated energy consumption arising from extensive computational operations poses a significant challenge. To address these issues, MEC has been introduced as a viable solution. By leveraging MEC, users at the network edge can offload resource-intensive computational tasks to nearby servers deployed at the network edge, thereby markedly reducing processing latency and energy consumption, thus alleviating the computational burden on end devices. Integrating UAVs with MEC systems offers a promising extension to traditional ground-based edge computing. UAVs are highly mobile and capable of establishing LoS communication links, allowing rapid and flexible network deployment independent of geographical constraints. UAV-MEC can provide on-demand computational support to ground users, particularly in remote, temporary, or disaster-stricken areas where fixed infrastructure is absent or inadequate. This enhances the coverage and responsiveness of MEC services, leading to a more adaptive and efficient resource allocation framework. However, the open and LoS nature of wireless channels in UAV-MEC systems makes the offloaded data highly susceptible to interception by malicious eavesdroppers. Transmissions between ground users and UAVs can be easily monitored, especially when unauthorized UAVs or other aerial agents act as eavesdropping nodes. Consequently, ensuring the confidentiality and integrity of user tasks during offloading becomes a major concern. Without proper security mechanisms, sensitive information is at risk of being exposed, highlighting the urgent need for strategies that can protect communication in such vulnerable settings. DRL has emerged as a powerful tool to address these complex challenges in UAV-MEC systems compared to traditional methods. DRL excels at handling sequential decision-making problems in unpredictable and dynamic environments, making it well-suited for optimizing task offloading, trajectory planning of UAVs, and anti-eavesdropping countermeasures. By leveraging its ability to learn and adapt from interactions with the environment, DRL provides an intelligent and scalable approach to improve both the efficiency and security of UAV-MEC networks.