MalFSM: Feature Subset Selection Method for Malware Family Classification

Malware detection has been a hot spot in cyberspace security and academic research. We investigate the correlation between opcode features of malicious samples perform feature extraction, selection fusion by filtering redundant features, thus alleviating dimensional disaster problem achieving efficient identification malware families for proper classification. authors use obfuscation technology to generate large number variants, which imposes heavy analysis burden on researchers consumes lot resources both time space. To this end, we propose MalFSM framework. Through method, reduce 735 contained Kaggle dataset 16, then fuse metadata (count file lines size) total 18 find that machine learning classification is high accuracy. analyzed interpreted selected features. Our comprehensive experiments show highest accuracy can reach up 98.6% only 7.76 s Microsoft.