Android malware detection with MH-100K: An innovative dataset for advanced research.

Area of Science:

• Computer Science
• Cybersecurity
• Machine Learning

Background:

• High-quality datasets are essential for effective supervised malware detection models.
• A significant challenge in machine learning for cybersecurity is the lack of representative and high-quality datasets.
• Existing datasets often fall short in providing comprehensive data for robust malware analysis.

Purpose of the Study:

• To introduce the MH-100K dataset, a large-scale collection of Android malware samples.
• To provide a public resource for the evaluation and comparison of machine learning-based malware classifiers.
• To facilitate research into Android malware prevalence, behavior, and evolution.

Main Methods:

• Compilation of 101,975 Android malware samples into the MH-100K dataset.
• Inclusion of detailed metadata in a CSV file: SHA256 hash, package name, API calls, permissions, and intents.
• Integration of VirusTotal analysis metadata for comprehensive sample information.

Main Results:

• The MH-100K dataset comprises 101,975 Android malware samples with extensive metadata.
• Metadata includes 166 permissions, 24,417 API calls, and 250 intents per sample.
• VirusTotal analysis data is incorporated, enabling deeper insights into malware characteristics.

Conclusions:

• The MH-100K dataset significantly enhances the availability of high-quality data for Android malware research.
• It supports advanced analysis of antivirus scan patterns and malware family behaviors.
• This resource is expected to advance the identification of new malware variants and the study of malware evolution.