The Impact of Federated Learning on IoT Security: An Empirical Analysis

This paper presents how Federated Learning (FL) will be applied in multiple areas, including smart homes, work environments, and healthcare systems, to enhance the security level of Internet of Things (IoT) systems. The rising number of Internet of Things devices has put focus on obscuring and managing these connected systems as a severe security issue. Traditional centralized classifiers frequently lack the required cyber-threat vulnerability, scalability, and data protection attributes. In this way, federated learning fills the communication gap between users (aka data producers) and servers by processing data locally, and on-device study demonstrates that the FL models get a significant advantage in the performance of the false positive reduction and the threat detection accuracy in comparison with the conventional models of operation. FL, though, may not be giving 92% accurate prediction in detecting the hazards as traditional models do with 88% correct prediction. Also, FL excelled in prior models by scoring 90% in industrial settings and 89% in the hospital setting, with an accuracy of 95% in the industrial setting and 93% in a hospital.Conversely, the recognition rate for ties grew from 5% to 3%, with 2% for correctly matched tie types compared to the traditional methods at 10% for correctly matched ties and 8% for corresponding tie types. The contribution is made using a mixed-methods approach comprising simulations and experimental sets to quantify the performance of FL. INTERPRETATION: FL has proved to be the best (from the results) among other security measures because it reduces data exposure and compliance with data privacy laws (strictness).