Industrial control anomaly detection data set equalization method based on constraint rule and similarity

The invention relates to the field of industrial control system safety, and discloses an industrial control anomaly detection data set balancing method based on constraint rules and similarity, which comprises the following steps: collecting sensor and actuator data of an industrial control system, forming an industrial control system anomaly detection training data set, training a twin network by using an unbalanced data set, and obtaining an anomaly detection result of the industrial control system. Making a data constraint rule according to the characteristics of the industrial control system, generating a random data sample according to the constraint rule, inputting the random data sample into the trained twin network to obtain a Hash vector of the sample, calculating the Hamming distance between the sample and all abnormal samples, and if the minimum distance is smaller than a threshold value, executing the next step; and if not, storing the random sample to the data set, and repeating the steps of rand