Identify Type of Android Malware with Machine Learning Based Ensemble Model
Recep Sinan Arslan
Abstract
The Android operating system is widely used in mobile devices thanks to its open source environment, flexible structure and features it offers. This situation makes it the target of cyber attackers and even worse, hackers are constantly developing their attack strategies in this area. Detecting and analysis of attacks in the Android OS presents a number of challenges due to security vulnerabilities and resource limitation on these mobile devices. In this study, an ensemble machine learning model is proposed to detect the type of malware (ransomware, adware, scareware or SMSmalware). The proposed model was trained and tested with the data in the CIC-AndMal-2017 dataset, which contains 4 malicious species. In the tests performed using 486 malicious sample, the malware type was detected with an accuracy of 90.4%. Precision, recall and precision values were also %90.4. It has been shown that ensemble models can yield better results than traditional classification algorithms in android malware type detection problems.