IoT terminal security vulnerability identification algorithm for new power systems

As the link node of multi-grid “fusion control” in energy interconnection, the safety of the new power system terminal is directly related to the safe and stable operation of the power grid. Once these terminals have security vulnerabilities and are maliciously exploited, it may lead to distortion of power grid monitoring data, tampering of control instructions, and even serious consequences such as power outages. In order to effectively ensure the stable operation of the new power system, an IoT terminal security vulnerability identification algorithm for the new power system is proposed. By combining graph embedding with pre-trained language models, complex relationships between IoT terminals are analyzed to extract security vulnerability features of IoT terminals. Using the gradient boosting decision tree (GBDT) model as the base classifier of the random forest, a security vulnerability recognition model for IoT terminals is constructed. The extracted features are used as inputs to the model, and the recursive feature elimination method is used to calculate the importance of the features until the feature subset is empty. The identification results of IoT terminal security vulnerabilities are obtained through collective voting. The experimental results show that the algorithm has a vulnerability recognition accuracy of 96.2%, a false positive rate of 0.5%, a false positive rate of 0.5%, and an AUC value of 0.987. Compared to the APAE algorithm, the accuracy has increased by 15.3% and the false positive rate has decreased by 82.1%. Compared to the CNN-GRU algorithm, the accuracy has increased by 12.7% and the false positive rate has decreased by 78.6%. When the number of terminals increased to 5000, the recognition accuracy remained at 97.5%, and the running time showed a linear increase, which can be adapted to large-scale IoT terminal deployment scenarios in new power systems.