A Robust Learning Framework for Smart Grids in Defense Against False-Data Injection Attacks

Abstract

With the rapid development of the application of smart grids in different sectors, security management has become a major concern due to cyber attack risks. Correctly and accurately estimating the real status of a smart grid under false-data injection attacks (FDIAs) is currently an emerging concern. In response to that concern, this work proposes a distributed robust learning framework for the inference of the working status under data integrity attacks. The proposed paradigm incorporates the technology median-of-means that enables identifying the correct state against various kinds of FDIAs that can efficiently prevent misleading information during the decision-making process in control centers. Compared with existing defense methods, our method is entirely data driven without training data, highly accurate, and reliable for wide-spectrum FDIAs. More important, it is capable of defending large-scale power electronic networks due to its distributed learning framework. Extensive experimental results demonstrate that our approach can provide efficient protection for Photovoltaic (PV) systems from FDIAs.