Signature Identification of False Data Injection Attacks Based on Deep Vision Networks
Yixuan He, Jingyu Wang, Dongyuan Shi · 2024
The increasing integration of cyber-physical systems has markedly amplified the vulnerability of power systems to False Data Injection Attacks (FDIAs), posing significant threats to their security and reliability. FDIAs can emulate various power system dynamics, each with distinct objectives, potentially leading to incorrect decisions by operators or causing damage that may go undetected initially. The impact of these attacks often hinges on their specific signatures, which reflect the temporal characteristics of the injected data waveforms. Therefore, precise classification of these attack signatures is essential for predicting their potential effects and devising effective countermeasures. This paper addresses this challenge by proposing the use of the VGG16 network, a deep learning model commonly utilized in computer vision, to classify the attack signatures. Leveraging VGG16's ability to identify complex patterns in data, the pro-posed method offers a robust solution for distinguishing between different types of FDIAs. Experimental validation on the IEEE 39-bus system confirms the effectiveness of this approach in improving the detection and classification of FDIAs.