A Modified FlowDroid Based on Chi-Square Test of Permissions

Area of Science:

• Computer Science
• Software Engineering
• Cybersecurity

Background:

• Android applications (apps) are prevalent in various fields, including IoT and embedded systems.
• Malicious apps exploit app permissions to compromise user privacy and data security.
• Existing static taint analysis tools like FlowDroid face challenges with complex detection and false alarms.

Purpose of the Study:

• To propose an improved static detection method for Android applications.
• To address the complexity and false alarm issues in existing taint analysis tools.
• To enhance the identification of dangerous data flows in Android apps.

Main Methods:

• Utilized Chi-square test to identify correlated permissions associated with malicious apps.
• Employed mutual information for clustering permissions into feature permission clusters.
• Developed a risk calculation method based on individual and combined permissions to detect hazardous data flows.

Main Results:

• The proposed method significantly enhances the efficiency of detecting dangerous data flows.
• Maintained high accuracy in identifying malicious data flows.
• Demonstrated a reduction in complexity and false alarms compared to FlowDroid.

Conclusions:

• The improved static detection method offers a more efficient and accurate approach to Android app security.
• Feature permission clustering and risk rating effectively identify dangerous data flows.
• This research contributes to more robust security analysis for Android applications.