Federated Learning for Privacy-Preserving Optimization in Multi-Domain Optical Networks: A Comprehensive Review

Abstract

Further sophistication of optical communication systems, coupled with increased speed of data traffic has necessitated the requirement of smart and scalable network management optimization techniques. Multi-domain optical networks are those networks in which various administrative domains are connected and work together, which bring various challenges in terms of privacy, data sharing, interoperability and resource optimization. Traditional centralized machine learning approaches are not going to be useful in such configurations since they need to aggregate the data, which can be a communication and privacy and security burden. The new paradigm of Federated Learning (FL) has emerged as a promising prospect to enable the joint training of models without accessing raw data. In this review paper, the idea of federated learning is discussed in detail in the context of multi-domain optical networks with a focus on how it could be used to implement privacy-preserving optimization. The paper discusses the fundamentals, structure, key algorithms, use case, issues, and future studies. It will equip the researcher and practitioners with an in-depth concept of how federated learning is able to revolutionize next-generation optical networking systems.