Intelligent Edge Computing for IoT: Enhancing Security and Privacy

Abstract

Edge computing is a distributed computing paradigm that involves processing data at or near the edge of the internet of things (IoT) network, instead of centralized server. This makes the cyber-attacks increasingly sophisticated, and traditional security measures become no longer sufficient to protect against them. Concurrently, privacy concerns arise when sensitive data is involved in Edge computing applications containing confidential information. In this paper, we propose a privacy-preserved federated learning (FL) approach for cyber-attack detection in edge based IoT ecosystem. A novel lightweight convolutional Transformer network (LCT) network is designed to precisely identify cyber-attacks though learning attack patterns from IoT traffics in local edge devices, where model is personalized though fine-tuning. The privacy of model and data is preserved in our system via incorporating differential privacy and secure aggregation during the cooperative training process on edge devices. We evaluate our proposed approach on a real-world dataset of network traffic (NSL-KDD) containing various types of attacks, and the experimental results show that our personalized FL approach outperforms traditional FL in terms of detection accuracy. We also show that our approach is effective in handling non-stationary data and adapting to changes in the network environment.