Integrating a Zero-Trust Adaptive Security Framework in Edge Based Federated Learning

Federated Learning (FL) allows multiple clients to collaboratively train a machine learning model without sharing their raw data, thus addressing privacy data concerns. The joint application of FL (Federated Learning) and MEC (Multi-Access Edge Computing) enables efficient utilization of computation and storage resources at the edge. By deploying the FL model at the edge, data are trained locally on the device or edge node, reducing latency, and facilitating real-time processing and decision-making, whilst also ensuring data privacy. However, the decentralized FL is subject to security and trust challenges, since MEC data may get poisoned. Therefore, ensuring the trustworthiness of clients is pivotal to ensure integrity and performance of FL models. This paper presents an adaptive Zero Trust framework which assumes no inherent trust in any client, is based on continuously verifying and validating the clients' behavior, to ensure that only reliable contributors (clients) are incorporated in the global model aggregation.