Detection of cyber attacks in smart grids using deep learning

Abstract

Abstract This paper explores the detection of False Data Injection (FDI) cyberattacks on smart grids using deep learning techniques. Cyberattacks can lead to undesirable outcomes, such as damage to electrical equipment, unnecessary relay tripping, or system-wide blackouts. The utilization of a Feature Tokenizer Transformer (FT-Transformer) model is proposed for robust detection accuracy compared to traditional tree-based classifiers. An FDI attack is mimicked on an IEEE 14 bus system to detect it using various machine learning algorithms. The proposed solution has also been validated against cyberattacks on a three-bus system. The results obtained demonstrate the superior performance of the FT-Transformer model in identifying FDI cyberattacks in smart grids. This research contributes to the advancement of modern cybersecurity power systems, ensuring their reliability against malicious cyber threats.