Android malware classification based on random vector functional link and artificial Jellyfish Search optimizer.

Area of Science:

• Computer Science
• Cybersecurity
• Machine Learning

Background:

• Android's dominant open-source market share makes it a prime target for malicious actors seeking user data.
• Existing Android malware detection relies on conventional machine learning, lacking metaheuristic optimization benefits.
• Malware exploits vulnerabilities for advertising, extortion, and data theft, necessitating advanced detection methods.

Purpose of the Study:

• To develop a novel Android malware detection system by optimizing Random Vector Functional Link (RVFL) models.
• To leverage the artificial Jellyfish Search (JS) optimizer for improved RVFL configuration and classification performance.
• To reduce the runtime of optimized malware detection models through dimensional reduction and metaheuristic optimization.

Main Methods:

• Dimensional reduction of Android application features.
• Optimization of Random Vector Functional Link (RVFL) models using the artificial Jellyfish Search (JS) optimizer.
• Evaluation on a dataset of 11,598 multi-class applications with 471 static and dynamic features.

Main Results:

• The RVFL+JS model demonstrated improved classification performance for Android malware detection.
• The optimization process led to minimized runtime for the detection models.
• The system effectively identified malicious applications within the tested dataset.

Conclusions:

• The proposed RVFL+JS system offers a more efficient and effective solution for Android malware detection.
• Metaheuristic optimization significantly enhances the performance of machine learning-based malware detection.
• This approach addresses the limitations of conventional methods by incorporating advanced optimization techniques.