Network Security Protection Model of Electric Power Information System Based on Hierarchical Weight Pruning Algorithm Optimizing OD-CNN Algorithm

The Internet of Things electric power information system is being used more and more frequently as a result of the advancement of IoT technology, and as a result, there is a higher chance that the network will try to eavesdrop on it. This research creates an Internet of Things intrusion detection model using a one-dimensional convolutional neural network and a hierarchical weight pruning technique to increase the protection level of the Internet of Things. The one-dimensional convolutional neural network model had the lowest accuracy and precision rate, and the other three models had only slight variations in recognition accuracy, according to the findings of the performance comparison studies. The recognition accuracy of the designed hybrid method, 1D convolutional neural network, Mobile Net, and Shuffle Net models is 92.6%, 90.7%, 91.2%, and 91.4% when the amount of test data is the entire test set, or 258. The designed hybrid algorithm, one-dimensional convolutional neural network, Mobile Net, and Shuffle Net models’ average intrusion detection speeds during the entire experiment were 18.2 ms, 53.6 ms, 24.3 ms, and 29.5 ms, respectively. When the computational samples were the entire test set, their memory consumption was 2893 KB, 18602 KB, 3741 KB, and 4262 KB. However, the computational speed is faster, the model is simpler, and it is suitable to be deployed in scenarios requiring fast feedback to identify the network intrusion detection results. It can be seen that the detection accuracy of the power information system IoT network intrusion detection model designed in this research has a negligible difference compared with the mainstream and novel models in the current market.