Integrating Federated Learning with Distributed Security for IoT-Driven Sports Technologies

Abstract

Increasing numbers of IoT devices and smart arenas along with athlete monitoring systems have formed data-rich environments in order to support high performance and minimize accidents. Centralized learning architectures can introduce delays in communication, security threats and privacy violations when dealing with sensitive data types like biometric or geo-location data. In this context, we propose the FedSecure-Sports framework, which solves these problems by introducing Walking in Federated Learning (FL) mechanisms and Distributed Security Mechanism to construct the sports ecosystems supported by IoT. The proposed algorithm employs the FedSecure Aggregation Algorithm (FSA-Alg), grounding on a light-weighted homomorphic encryption, blockchain consensus and weighted federated model updates. This enables the training of a model over remote devices without sharing raw data, verifying updates, and imposes security against tamperings by malicious attackers. FedSecure-Sports achieves superior anomaly detection accuracy as well the ability to reduce communication cost (by 15%) and protect privacy (18-22%) compared with FL and centralized methods on standard IoT sports datasets. These findings support that our solution improves the reliability, robustness, and scalability of realtime sports analysis. In conclusion, FedSecure-Sports proposes a safe, privacy-protecting, and efficient IoT-based sports technology future.