Related Experiment Video
Updated: Jan 7, 2026
08:20
Author Spotlight: AI-Driven Trypanosome Species Detection from Microscopic Images
Published on: October 27, 2023
2.4K
MaSS-Droid: Android Malware Detection Framework Using Multi-Layer Feature Screening and Stacking Integration
Zihao Zhang1,2, Qiang Han1,2, Zhichao Shi1,2
1School of Computer Science and Engineering, North Minzu University, Yinchuan 750021, China.
Entropy (Basel, Switzerland)
|December 24, 2025
Summary
This study introduces MaSS-Droid, a novel Android malware detection framework. It enhances security by reducing feature redundancy and improving model stability and accuracy for complex threats.
Area of Science:
- Computer Science
- Cybersecurity
- Machine Learning
Background:
- Android malware is a growing threat, challenging user security.
- Malware detection faces issues with feature redundancy and unstable single model performance.
Purpose of the Study:
- To propose MaSS-Droid, a framework for robust Android malware detection.
- To address feature redundancy and improve ensemble model generalization and stability.
Main Methods:
- Extracted permission, API call, and opcode sequence features from APK files.
- Implemented a three-layer feature screening mechanism to reduce redundancy and complexity.
- Utilized an adaptive Stacking integration method (Adaptive-Stacking) to dynamically adjust base classifier weights.
Main Results:
- MaSS-Droid effectively mitigates overfitting and enhances model generalization.
- The framework significantly reduces feature redundancy, improving detection accuracy.
- Demonstrated enhanced overall stability and accuracy in detecting diverse Android malware samples.
Conclusions:
- MaSS-Droid offers a stable and accurate solution for Android malware detection.
- The proposed framework effectively tackles challenges posed by feature redundancy and model instability.
- Adaptive Stacking integration proves crucial for superior ensemble performance in malware analysis.
