Generative Edge Intelligence for Securing IoT-assisted Smart Grid against Cyber-Threats

Abstract

The critical dependence of industrial smart grid systems on cutting-edge Internet of Things (IoT) technologies has made these systems more susceptible to a diverse array of assaults. This consequently puts at risk the integrity of energy data as well as the safety of energy management activities that depend on those data. This study offers a generative federated learning framework for semi-supervised threat detection in an IoT-assisted smart grid system. We refer to this framework as FSEI-Net. A unique semi-supervised edge intelligence network (SEI-Net) is presented in the FSEI-Net to enable semi-supervised training using labeled and unlabeled data in the edge tier. The design of SEI-Net is based on with bidirectional generative convolutional network that can intelligently capture the patterns of threat data from partially labeled smart grid data. We present federated training to enable remote edge servers to work together on training a semi-supervised detector without disclosing their own private local data. This is accomplished through cooperative training. To facilitate communication between cloud and edge layers that is both secure and respectful of users' privacy, a reputation-based block chain is introduced in the FSEI-Net. The outcomes from the practical applications demonstrate that the effectiveness of the proposed FSEI-Net over the most recent cutting-edge detection approaches are valid