Abstract
The rapid evolution of malware and cyber threats has made traditional centralized detection systems less effective due to privacy concerns, data silos, and limited adaptability. To address these challenges, this study proposes Cyber Force, a federated reinforcement learning-based framework for intelligent malware detection. The system leverages federated learning to enable multiple distributed devices or organizations to collaboratively train models without sharing sensitive data, thereby preserving privacy while improving detection performance. Reinforcement learning is integrated to dynamically adapt detection strategies based on environmental feedback, allowing the system to continuously improve its decision-making capabilities. The proposed framework combines decentralized data training with adaptive learning mechanisms to detect both known and unknown malware variants effectively. By utilizing local training on edge devices and aggregating model updates through a central server, the system ensures scalability and robustness. Experimental analysis demonstrates that the Cyber Force framework achieves higher detection accuracy, reduced communication overhead, and enhanced privacy compared to traditional centralized approaches. This research contributes to the development of next-generation cybersecurity systems capable of handling evolving malware threats in distributed environments.