Abstract
With the rapid growth and development of smart cities, many real-time systems have been designed to support the dynamic infrastructure of realistic and time-constrained networks. These systems integrate the Internet of Things (IoT), emerging technologies, and edge computing to meet the demands of diverse applications and fulfill community needs. Researchers have made significant efforts to develop autonomous systems for manipulating network data, providing timely responses in critical environments, and easing communication by enabling seamless device-to-device interaction. However, some significant research challenges still persist, including the lack of intelligence and security over unpredictable communication under malfunctioning conditions. Moreover, uneven load balancing across devices with continuous link interference results in network disconnection and delays in system responsiveness. In addition, the devices' heterogeneity and unreliable links make most existing edge-driven schemes complicate the process of optimal offloading. To address these issues, we introduce a joint optimization framework for energy-efficient, secured computation offloading and dynamic resource allocation in Edge-IoT infrastructure. It explores a Multi-Agent Reinforcement Learning to make adaptive offloading decisions and establishes a secure communication path with effective resource allocation under dynamic network interactions. Furthermore, it enhanced network stability by leveraging trusted, decentralized processing points at the network edge, mitigating energy holes, and improving workload management for distributed IoT ecosystems. Using an extensive set of simulations, the proposed framework is evaluated, and the results reveal substantial improvements over recent state-of-the-art solutions in terms of both security and energy efficiency.