Android Malware Detection Using Kullback-Leibler Divergence

• Main Authors: Vanessa N. COOPER, Hisham M. HADDAD, Hossain SHAHRIAR
• Format: Article
• Language: English
• Published: Ediciones Universidad de Salamanca 2015-03-01
• Series: Advances in Distributed Computing and Artificial Intelligence Journal
• Subjects: android malware detection; kullback-leibler divergence; back-off smoothing
• Online Access: https://revistas.usal.es/index.php/2255-2863/article/view/12296

Many recent reports suggest that mareware applications cause high billing to victims by sending and receiving hidden SMS messages. Given that, there is a need to develop necessary technique to identify malicious SMS operations as well as differentiate between good and bad SMS operations within applications.<br />In this paper, we apply Kullback-Leibler Divergence (KLD) as a distance metric to identify the difference between good and bad SMS operations. We develop a set of elements that represent sending or receiving of SMS messages, both legitimately and maliciously. Then, we compare the divergence of the trained set of elements. Our evaluation shows that the divergence between good and bad applications remains significantly high, whereas between two applications performing the same SMS operations remain low. We evaluate the proposed KLD-based concept for identifying a set of malware applications. The initial results show that our approach can identify all known malware and has less false positive warning.