Energy-Aware Adaptive Federated Cyber Defense for Internet of Drones using Multi-Objective Reinforcement Learning

The Internet of Drones (IoD) enables cooperative aerial applications but introduces significant cybersecurity risks due to open wireless communication, distributed coordination, and limited onboard energy resources. Existing federated learning–based security approaches emphasize privacy but lack energy-aware and adaptive defense mechanism. This paper proposes an Energy-Aware Adaptive Federated Cyber Defense (EA-AFCD) framework that integrates lightweight local intrusion detection with customized federated learning and a multiobjective reinforcement learning controller. The controller dynamically manages drone participation, aggregation frequency, and model update strategies by jointly optimizing detection accuracy, energy consumption, and communication overhead. A non-IID federated environment is simulated using a benchmark intrusion detection dataset to reflect realistic IoD conditions. Experimental results show that EA-AFCD achieves higher detection accuracy 98.5% significantly reducing energy usage and communication cost. The framework also extends network lifetime and improves learning convergence. The findings demonstrate that adaptive, energy-aware federated defense enhances scalability and practicality for real-world IoD deployments.