Hybrid deep learning model for accurate and efficient android malware detection using DBN-GRU.

Area of Science:

• Computer Science
• Cybersecurity
• Artificial Intelligence

Background:

• Android applications face increasing security threats from advanced malware.
• Traditional detection methods struggle against polymorphic and metamorphic malware variants.

Purpose of the Study:

• To develop a hybrid deep learning model (DBN-GRU) for enhanced Android malware detection.
• To improve detection accuracy and efficiency against sophisticated malware.

Main Methods:

• A hybrid deep learning model (DBN-GRU) integrating Deep Belief Networks (DBN) for static analysis and Gated Recurrent Units (GRU) for dynamic analysis.
• Extraction of static features (permissions, API calls, intent filters) and dynamic features (system calls, network activity, inter-process communication) from Android APKs.
• Training and testing on the Drebin dataset (129,013 applications).

Main Results:

• DBN-GRU achieved 98.7% accuracy, 98.5% precision, 98.9% recall, and an AUC of 0.99.
• Outperformed conventional models like NMLA-AMDCEF, MalVulDroid, and LinRegDroid.
• Demonstrated faster preprocessing, feature extraction, and classification times, suitable for real-time deployment.

Conclusions:

• The DBN-GRU model effectively bridges static and dynamic detection, enhancing malware detection capabilities.
• The model reduces false positives and computational overhead, offering a scalable solution.
• Confirmed applicability in real-world Android security for high-performance malware detection.