A New Approach for Federated, Shared, and Collaborative Learning via a SAN Storage Network

Abstract

The exponential emergence of IoT devices in a smart city generates significant data traffic on the network. This traffic can be affected by changes in the values exchanged between IoT devices and the aggregation server, which further guarantees the lack of parameter integrity, leading to unreliable learning. For this reason, we propose a new approach to protecting the private and confidential data of IoT devices required for federated learning of an AI model. Our approach aims to improve accuracy of learning model predictive values based on Fiber Channel technology, which adjusts the sharing and storage of learning parameters over the network, ensuring a reliable connection between IoT devices and storage arrays. This technology accelerates the AI model training phase while reducing latency.