期刊名称：IAENG International Journal of Computer Science
印刷版ISSN：1819-656X
电子版ISSN：1819-9224
出版年度：2019
卷号：46
期号：4
页码：524-533
出版社：IAENG - International Association of Engineers
摘要：Android continues to dominate the mobile socialdevices, and Android applications have become the major targetof hackers in social networks. Although millions of Androidmalware samples are found every year, they can be groupedinto a limited number of malware families. To automaticallyand effectively classify Android malware into the correspondingmalware families, a deep-learning based classification approachis proposed by utilizing the code-images converted from themalware’s binary bytecodes. To overcome the training issuethat only a very limited amount of malware samples arepublicly labeled with families, the deep-learning classifier makesuse of the low-level layers of a pre-trained convolutionalneural network. The empirical studies show that the proposedapproach excels the existing code-image based technique inimplementation simplicity as well as in classification metricssuch as F-measure values, false positive rates, and false negativerates. Furthermore, the implemented classifier can identifymalware families of different sizes, including small families.
关键词：Android malware; code image; deep learning;malware family classification
