A fuzzy classification method based on feature selection algorithm in malicious script code detection

In this paper, a new feature selection algorithm was uesd in fuzzy classification to detect malicious script code. Firstly, extract statistical features from samples based on knowledge and the key words. Next, correlation and separability criterion based on minimum mean square error was utilized to filter the features. After that, the noise of samples was deleted by the final features. The features matrix of script samples from the malicious script collection and the benign script collection was obtained. As to fuzzy model, normal distribution of partial large-scale was selected to construct the membership function according to the malicious script features. The results of experiment show that the fuzzy classification method using feature selection algorithm has higher accuracy than the method using variance feature selection.