SAMLDroid: A Static Taint Analysis and Machine Learning Combined High-Accuracy Method for Identifying Android Apps...

Area of Science:

• Computer Science
• Cybersecurity
• Machine Learning

Background:

• Insecure applications pose a growing threat by stealing user location data for illicit activities.
• Existing static and dynamic taint analysis methods for identifying such apps have limitations, including false positives and negatives.

Purpose of the Study:

• To develop a more accurate method for identifying Android applications with location privacy leakage.
• To reduce misjudgment rates in vetting suspicious applications.

Main Methods:

• Proposes SAMLDroid, a hybrid approach combining static code analysis and machine learning.
• Static analysis scrutinizes source code for location-acquiring intentions.
• A trained machine learning classifier analyzes app features for dynamic pattern detection.

Main Results:

• SAMLDroid achieves up to 98.4% accuracy in identifying apps with location privacy leakage.
• Demonstrates a significant improvement in identification rates compared to existing methods like Apparecium.

Conclusions:

• SAMLDroid effectively improves the accuracy of detecting Android apps with location privacy leakage.
• The combined static and machine learning approach offers a superior solution for mobile application security vetting.