Novel Multi-Classification Dynamic Detection Model for Android Malware Based on Improved Zebra Optimization Algorithm...

Area of Science:

• Cybersecurity
• Machine Learning
• Mobile Security

Background:

• Android malware is rapidly increasing, posing a significant threat to smartphone users.
• Current static analysis methods struggle with sophisticated obfuscation techniques used by malware.
• There is a need for more effective and accurate Android malware detection methods.

Purpose of the Study:

• To propose a novel dynamic detection model for Android malware.
• To enhance the accuracy and efficiency of Android malware detection.
• To address the limitations of static analysis in detecting obfuscated malware.

Main Methods:

• Developed an Improved Zebra Optimization Algorithm (IZOA) with elite opposition-based learning and firefly perturbation.
• Utilized IZOA to optimize hyperparameters for the Light Gradient Boosting Machine (LightGBM) model.
• Implemented a dynamic detection model, IZOA-LightGBM, for multi-classification of Android malware.

Main Results:

• The IZOA-LightGBM model achieved high detection accuracies: 99.75% on CICMalDroid-2020, 98.86% on CCCS-CIC-AndMal-2020, and 97.95% on CIC-AAGM-2017.
• The proposed model demonstrated superior performance compared to other existing models.
• Enhanced convergence speed and search capability of the optimization algorithm.

Conclusions:

• The IZOA-LightGBM model offers a highly effective solution for dynamic Android malware detection.
• The integration of IZOA and LightGBM significantly improves detection accuracy against complex malware.
• This approach provides a robust defense against the growing threat of Android malware.