Research on Industrial Control Network Security Data Feature Extraction Technology Based on Composite Sparse Autoencoder

Nowadays, industrial control system networks generate massive network communication data during their operation, characterized by large volume, diverse attributes, significant noise, rapid changes, and strong sparsity. Effectively and rapidly identifying network attack features from this data poses a challenging problem. This paper explores a dimensionality reduction model for network security big data, conducts an analysis of network security data features under the context of big data, and proposes a method for extracting industrial control network security data features based on a composite kernel sparse autoencoder. This method proves to be effective in enhancing the recognition rate of network security feature extraction.