Federated Learning for Decentralized Cyber Defense: A Systematic Review of Emerging Trends and Open Challenges

Abstract

The exponential growth of data and the rapid evolution of Artificial Intelligence have reshaped cybersecurity by enabling more sophisticated threat detection and defense mechanisms. Traditional Machine Learning models rely on centralized data aggregation for training, which increases data exposure risks and raises significant privacy and security concerns. Federated Learning offers a decentralized paradigm that allows collaborative model training across distributed devices while keeping sensitive data local. This approach enhances privacy and compliance with data protection regulations, but it introduces new challenges, including data heterogeneity, communication costs, and privacy preservation. These challenges limit the large-scale and reliable adoption of Federated Learning in cyber defense systems. To address these challenges, recent studies have explored integrating complementary technologies, such as Blockchain for secure coordination, quantum computing for computational optimization, and advanced neural architectures for adaptive learning. This study systematically reviews current research, identifies persistent challenges, and discusses future directions for advancing decentralized cyber defense through Federated Learning.