DDoS Attack Detection in Smart Grid Networks Using Visual Features

Smart grid networks play a key role in making electric power infrastructure more efficient and ensuring reliable energy distribution. However, the interconnected communication devices in these networks expose the system to several network threats, such as the distributed denial of service (DDoS) attacks. Timely DDoS attack detection and mitigation is an important but challenging task for maintaining service integrity. In this paper, we propose to leverage computer vision models to detect DDoS attacks in smart grid networks. After preprocessing the input network data and extracting features, we transform the features into two-dimensional images, employing them to train and fine-tune a vision MLP mixer model. Our approach offers the advantage of adaptability, as any pre-existing computer vision model can be utilized for fine-tuning purposes, providing flexibility in the absence of readily available pretrained models in the DDoS detection domain. Performance metrics, including accuracy, precision, recall, and F1 score obtained from our experiments, show promising results for DDoS attack detection in smart grid networks using the UNB-ISCX-IDS dataset, which contains real-world network traffic data containing different attack scenarios. Our approach will be beneficial for robust and secure smart grid operation in real-world scenarios.