Android Malware Analysis using Coefficient of Multiple Correlation
Shresth Jain, Sarthak Kapoor, Anshul Arora, Sachin Kumar Sharma · 2023
Android being the most famous Operating System (OS) for smart hand-held devices also serves as a prime attraction for cyber-criminals and black hat hackers. Hacking into these devices through malware applications gives them access to the data which can be used for personal gains. These applications have been a prime source of cyber crimes, information leaks, financial frauds, and much more. In order to get control over malware applications, it is of utmost importance to detect these applications before they are installed on any system. This can prevent huge losses for mankind. Malware detection has become a challenge due to constant technological advancements. Most of the data about an app including permissions, activities, services, etc can be extracted using its manifest file.The concept of Multiple Correlation Coefficient has been used to rank these permissions and then build a machine learning and deep learning model using various classifiers in a k-fold setup. The results suggest that the presented model gives a detection accuracy of 97.54% with Random Forest Classifier. This accuracy is observed with the top 210 permissions as ranked by Multiple Correlation Coefficient.