Enhancing Android Malware Detection through Swarm Intelligence and Autoencoder Techniques Applied to API Call...

Area of Science:

• Cybersecurity
• Machine Learning
• Software Engineering

Background:

• Malware security intelligence relies on analyzing application data for threats.
• Application Programming Interface (API) calls are crucial for malware detection.
• Reducing feature space improves malware analysis efficiency.

Purpose of the Study:

• To pinpoint significant API call features for precise Android malware detection.
• To enhance the efficiency and accuracy of malware identification systems.

Main Methods:

• Utilized swarm intelligence optimization techniques: Firefly Optimization, Cuckoo Search, and Ant Colony Optimization.
• Employed Auto-Encoders for significant feature extraction.
• Evaluated feature sets using machine learning classifiers (KNN, RF, SVM, DT, LR) and a hybrid neural classifier.

Main Results:

• Achieved a high accuracy of 98.87% in malware categorization.
• Demonstrated effectiveness using only 7 out of 100 API call features.
• A hybrid artificial neural classifier improved malware categorization performance.

Conclusions:

• Swarm intelligence effectively identifies critical API call features for Android malware detection.
• The proposed method significantly enhances detection precision and efficiency.
• This approach offers a robust solution for improving mobile security intelligence.