ML-Based Ransomware Detection for Android OS

Abstract

Nowadays, there are billions of users of Android smartphones. Smartphones are acting like small computers. It can help scientists in many ways, such as bird monitoring, earthquake sensing, tree identification, data collection, cloud computing, and many others. Other users, can browse, collect information, and save it on the phone. As Android operating system usage grows, the security stakes are rising. Some of the information is considered personal and sensitive data. The hackers or the malware authors exploit the vulnerabilities on the OS to get the sensitive data of the users. Ransomware is used to encrypt the data and sometimes lock the phone. The open-source feature on Android is one of these vulnerabilities, so Android is more likely to be attacked. Studies provided some techniques based on machine learning to detect ransomware, but it does not avoid it and it still reaching. In this project, we will discuss a machine learning-based technique for Android ransomware detection. The proposed model will be mainly based on attributes related to system calls generated by the application and LogCat logs.

After extracting features from System Calls and LogCat logs, we will Classify the applications using several machine learning classifiers to test the model performance. The results show that ensemble learning has the best detection accuracy of 0.94 using the Random Forest algorithm.

Date of Award	Apr 2023
Original language	American English
Supervisor	Chan Yeun (Supervisor)