A Fast Approach Towards Android Malware Detection

摘要

The proposed research compares the feasibility of three well known machine learning algorithms on the detection of malware on the Android platform. Once accuracy is at an acceptable level, these algorithms performance are further enhanced to decrease analysis time, which can lead to faster detection rates. The framework makes use of powerful GPU's (Graphics Processing Unit) in order to reduce the time spent on computation for malware detection. Utilizing MATLAB's parallel computing kit, we can execute analysis at a much higher speed due to the increased cores in the GPU. A reduced computation time allows for quick updates to the user about zero day malware, resulting in a decreased impact. With the increase in mobile devices unending, quick detection will become necessary to combat mobile malware, and with Android alone reaching its 50 billionth app downloads will be no small task.