Abstract
In this paper, a privacy-preserving malware detection framework for cloud infrastructures is presented using federated learning and deep neural networks. The proposed system is designed to address the growing challenges of data privacy, distributed cloud environments, and evolving malware threats. Federated learning is employed to enable collaborative model training across multiple cloud clients while ensuring that sensitive data remain on local devices and are never centrally shared. Advanced deep learning (DL) architectures, namely Transformer-based and GRU-enhanced models, are integrated to capture both long-range contextual dependencies and temporal behavioral patterns in cloud network traffic. An ensemble learning strategy is adopted to combine the strengths of both models, resulting in improved detection accuracy and robustness. The system is evaluated using a synthetic dataset that simulates real-world cloud traffic scenarios, and performance is assessed through standard metrics including accuracy, precision, recall, F1-score, AUC, and confusion matrix analysis. Experimental results demonstrate that the proposed approach achieves high detection performance while maintaining strong privacy guarantees. The findings confirm that the integration of federated learning with hybrid DL models provides a scalable and effective solution for secure malware detection in cloud environments, making the framework suitable for deployment in privacy-sensitive and large-scale distributed systems.