Mobile edge computing (MEC) improves the performance of real-time services by bringing computing resources closer to users, thus reducing latency and optimizing network utilization. Task offloading plays a key role in this architecture, intelligently distributing loads between edge devices and servers, guaranteeing efficient use of resources and improved energy efficiency. This decentralization exposes edge infrastructures to increased cyber threats, making it essential to integrate intrusion detection systems (IDS) to guarantee the security and reliability of MEC environments. In this paper, we present an IDS model that uses the Deep Q-Network (DQN) algorithm with the ensemble learning method, specifically the voting classifier that combines three algorithms Random Forest (RF), Support Vector Machine (SVM) and Gradient Boosting (GB) on the two datasets Shanghai and Guangzhou. The model gave an accuracy between 67 and 99%.