Federated learning (FL) has emerged as a powerful paradigm in cyber security research, enabling collaborative model training across distributed devices and organizations without the need to share raw data, thereby preserving privacy while strengthening defense against evolving threats. Research papers in this field investigate the use of FL for intrusion detection systems, malware detection, phishing defense, and anomaly-based threat monitoring in domains such as IoT, cloud, and edge networks. Studies emphasize its ability to leverage decentralized data sources for more accurate threat intelligence, while mitigating risks associated with data leakage and centralized failures. Recent works also explore challenges such as communication overhead, heterogeneous data distributions, poisoning attacks, and adversarial manipulation in federated settings. To address these, researchers propose secure aggregation, differential privacy, blockchain integration, and robust aggregation algorithms. Federated learning for cyber security is increasingly seen as a promising solution to build scalable, privacy-preserving, and collaborative defense mechanisms that adapt to the dynamic landscape of modern cyber threats.