Decentralized Federated Learning for Internet of Things Anomaly Detection

Abstract

With the improvement of computing power and the development of network technology, Internet of Things (IoT) devices are widely used in many industries. But it also faces various security threats. Anomaly detection is a commonly used method, but traditional methods face shortcomings such as low accuracy. Therefore, in this paper, we introduce a decentralized federated learning method for anomaly detection, using neural networks to improve accuracy and take advantage of the characteristics of federated learning to protect local data security. The decentralized algorithm avoids the drawbacks of traditional federated learning such as the single point of failure. Finally, we conduct simulation experiments on the IoT23 dataset, which verify the performance of our system.