Android Malware Classification Addressing Repackaged Entities by the Evaluation of Static Features and Multiple Machine Learning Algorithms

Abstract

Expanded usage and prevalence of android apps allows developers of malware to create new ways in various applications to unleash malware in various packaged types. This malware causes various leakage of information and a loss of revenue. In addition, the discovered software is repeatedly launched by unethical developers after classifying the program as malware. Unluckily, the program still remains undetected even after being repackaged. In this research, the topic of repackaging was discussed, emphasizing the implementation based on source code using the bag-of-words algorithm and testing the findings through machine learning. The findings of the assessment demonstrate comparatively improved result in this aspect than the existing implantation based on source code by adapting the bag-of-words strategy and implementing some supplementary dataset preprocessing. A vocabulary for identifying the malicious code has been developed in this study. Bag-of-words was used to classify malware trends using custom implementation. The findings were instantiated using various algorithms of machine learning. The concept was eventually implemented in a practical application too. The suggested method sets out a fairly new methodology for examining source code for android malware to tackle repackaging of malware.