Software vulnerability assessment method and system based on comparative learning and medium

The invention discloses a software vulnerability assessment method and system based on comparative learning and a medium, and the method comprises the steps: employing a training method based on comparative learning to train a feature extraction model, the training method of comparative learning comprises the following steps: learning and extracting feature vectors of vulnerability attributes by comparing positive samples and negative samples of the vulnerability attributes; extracting feature vectors from the vulnerability attributes of the to-be-evaluated vulnerability by using the trained feature extraction model; and classifying the feature vectors by using a vulnerability classifier to obtain a classification result of the to-be-evaluated vulnerability. According to the method, more effective high-dimensional vulnerability feature representation is obtained by utilizing comparative learning in the feature extraction process, so that the accuracy of software vulnerability assessment is improved. 本发明公开了一种基