Abstract: Researchers give preference to dynamic analysis based malware detection methods with capability of nullifying the effects of polymorphism and obfuscation on malware and detecting new and unseen malwares.In this case,malware authors embed numerous anti-detection functions in to malware to evade the detection of existing dynamic malware detection methods.To solve this problem,a malware detection method MACSPMD based on malicious API call sequential pattern mining was proposed.Firstly,dynamic API call sequences of the files are gotten by real machine which simu-lates the actual running environment of the malware.Secondly,the malicious API call sequence patterns that can represent the potential malicious behavior patterns are mined by introducing the concept of objective-oriented association mining.Finally,the malicious API call sequences are used as abnormal behavior feature to detect malware.The experimental results based on real data set show that MACSPMD achieves 94.55% and 97.73% of detection accuracy on unknown and evasive malware respectively.Compared with other malware detection methods based on API call data,the detection accuracy of unknown and evasive malware is improved by 2.47% and 2.66% respectively,and the time consumed in the mining process is less.MACSPMD can effectively detect known and unknown malware,including escape type.

Key words: Malware detection,Evasive-malware,Sequetial pattern mining,API call sequence,Classification