A static analysis approach for Android permission-based malware detection systems.

Area of Science:

• Computer Science
• Cybersecurity
• Machine Learning

Background:

• Mobile malware evolution necessitates robust detection methods.
• Current security evaluations for Android malware, including static and dynamic analysis, show room for improvement.

Purpose of the Study:

• To evaluate the effectiveness of static analysis for Android malware detection using permission-based features.
• To compare the performance of various machine learning classifiers for identifying Android malware.

Main Methods:

• Utilized a dataset of 5,000 Drebin malware and 5,000 Androzoo benign samples.
• Employed machine learning with feature selection to identify distinguishing malware characteristics.
• Compared the performance of different classifier sets for Android malware detection.

Main Results:

• The Random Forest algorithm demonstrated the highest accuracy in malware detection.
• Achieved a True Positive Rate (TPR) of 91.6% using the Random Forest algorithm.
• Identified key permission-based features crucial for distinguishing malware from benign applications.

Conclusions:

• Static analysis, enhanced by machine learning and feature selection, is effective for Android malware detection.
• The Random Forest algorithm offers a highly accurate solution for identifying Android malware.
• Permission-based features are valuable indicators for improving mobile security evaluations.