Classification of malicious software behaviour detection with hybrid set based feed forward neural network

Abstract

Behavior detection of malicious software is better than signature-based detection method when used to find unknown malicious software. The paper presents a classification method of malicious software behavior detection with hybrid set based feed forward neural network. We choose malicious software detection database for test with 57345 records from National Anti-Computer Intrusion and Anti-Virus Research Center. According to the definition of selected data set relations and transfer functions, the weighted path length trees of malicious software detection data are calculated for neural network input vectors. After repeat training, different malicious software detection methods can be classified by the method with the about 83.9 percent right classification. © 2010 Springer-Verlag.