Federated Learning in Edge Computing: Advancements, Security Challenges, and Optimization Strategies

Federated Learning (FL) has emerged as a transformative paradigm in edge computing, enabling decentralized model training across distributed devices while preserving data privacy. Unlike traditional machine learning (ML) approaches that rely on centralized data aggregation, FL facilitates on-device learning, reducing data transfer overhead and enhancing security. However, conventional FL implementations encounter challenges such as communication inefficiencies, model convergence issues, and susceptibility to adversarial attacks. This research analyses challenges and explores strategies to improve FL’s efficiency, security, and scalability in edge environments. Additionally, the heterogeneity of edge devices and network constraints complicates FL deployment in real-world applications. This research advances FL in edge computing by addressing these limitations through optimized communication strategies, enhanced security mechanisms, and adaptive aggregation techniques. The research explores various methodologies to improve model performance while ensuring robust privacy preservation in resourceconstrained environments. Furthermore, it evaluates FL’s security vulnerabilities in edge networks and proposes mitigation strategies against threats such as data poisoning and model inversion attacks. Artificial intelligence (AI) advancements enhance FL through optimization, adaptive learning, and privacy-preserving models. Innovations like federated reinforcement learning and automated tuning improve efficiency, robustness, and adaptability in diverse edge networks. By contributing to the development of more efficient and secure FL architectures, this research provides valuable insights into improving model generalization, reducing latency, and enhancing resilience against cyber threats. These advancements pave the way for scalable and privacy-preserving AI solutions, reinforcing the role of $F L$ in the future of edge computing.