AndroDex: Android Dex Images of Obfuscated Malware.

Area of Science:

• Computer Science
• Cybersecurity
• Machine Learning

Background:

• Increasing cyber threats necessitate robust Android malware detection.
• Existing malware datasets may not adequately address obfuscation techniques.
• A large-scale, reliable dataset is crucial for developing effective malware classifiers.

Purpose of the Study:

• To introduce AndroDex, a novel Android malware dataset.
• To provide a resource for enhancing Android malware detection and classification research.
• To facilitate deep learning-based malware analysis.

Main Methods:

• Collected 24,746 Android malware samples across 180+ families.
• Converted malware Portable Executable (.dex) files into image formats.
• Applied obfuscation techniques to malware samples prior to image conversion.

Main Results:

• AndroDex dataset comprises 24,746 malware samples represented as images.
• The image-based representation effectively visualizes malware characteristics, bypassing obfuscation.
• The dataset supports static analysis of .apk files, offering a time- and space-efficient initial examination.

Conclusions:

• AndroDex is a valuable resource for advancing Android malware detection and classification.
• Malware image analysis offers a promising approach to overcome obfuscation challenges.
• The dataset facilitates initial static analysis of .apk files, aiding malware research.