Malware detection for mobile devices

• Main Author: Wei Hao, Lew
• Other Authors: Lin Shang-Wei
• Format: Final Year Project
• Language: English
• Published: 2018
• Subjects: DRNTU::Engineering
• Online Access: http://hdl.handle.net/10356/74229
• Institution: Nanyang Technological University

This paper describes the techniques used to detect Android malware using a machine learning approach. Even though the title of the paper is mobile devices, this paper only focuses on the Android smartphone. This paper starts off by providing some of the background knowledge that is related to this field. Then, it proceeds to discuss some of the experimentation processes which talks about how and why certain approaches are selected over others during the project. It then goes into greater details about the implementation. It then goes on to discuss the evaluation results before concluding the discussion by talking about how this project could be further improved as well as how this project could be applied in the real world context. Some of the contributions of this paper includes using new features for machine learning that have not been explored by other papers before. Some of these new features include cyclomatic complexity and ngrams constructed from sensitive sources to sinks. Most research paper trains model according to each specific malware family. This paper takes a slightly different approach by aggregating similar malware families together and train a model for the new aggregated family instead. The advantage of this is that there is a larger dataset, which leads to higher reliability. This paper also contributes the results in terms of the accuracies for each classifier and malware family, as well as the most significant features for each malware family.