Abnormal flow detection using Optimized ABFlow Network in SDN based Smart Grid Application

SDN can help the power communication network function more efficiently and conform to the Smart Grid's need for centralised control. There are a variety of network attacks that might be launched against the SDN controller. A significant danger to the data integrity of smart grids is posed by malicious software, which often use encryption or tunnelling techniques to get beyond firewalls, intrusion detection systems, and other protective measures. For the safety and reliability of the Smart Grid, it is crucial that irregular flow be detected accurately. Conventional machine learning techniques, such as and the Naive Bayes classifier, were the basis of earlier efforts. Low accuracy for huge, high-dimensional network flows are a result of their simplistic, surface-level feature learning. In this study, we create a system for detecting anomalous flows, called ABFlow, using Optimized Siamese neural networks. These networks operate well when only little data is provided for training. By examining the trajectory data with several parameter measurements, the model may identify irregular traffic flows. The suggested strategy is then tested on the DDoS-SDN and InSDN datasets to determine how well it performs. The experimental findings prove that the ABFlow can effectively identify anomalous flow in the SDN-based Smart Grid, greatly outperforming existing tactics in terms of accuracy and FPR.