YATSIDroid: an android malware detection framework based on artificial immune system.

Area of Science:

• Software Engineering
• Cybersecurity
• Machine Learning

Background:

• The exponential growth of Android apps presents significant malware detection challenges.
• Cybercriminals exploit the Android permission model to distribute malware through deceptive app repositories.
• Effective malware detection is crucial for protecting smartphone users.

Purpose of the Study:

• To propose a novel malware detection model for Android applications.
• To enhance the accuracy of identifying malware-infected apps, especially with limited labeled datasets.
• To leverage feature ranking and a two-stage semi-supervised classifier for improved detection.

Main Methods:

• Development of YATSIDroid, a meta-algorithm combining feature ranking and a two-stage semi-supervised classifier (Yet Another Two-Stage Idea - YATSI).
• YATSI allows distinct machine learning algorithms in its first stage.
• Utilized a limited labeled dataset for training and evaluation.

Main Results:

• YATSIDroid demonstrated improved performance in identifying malware-infected apps.
• The model achieved a 98.5% detection rate for malware-infected apps.
• Effective detection was accomplished using only 60% of the labeled dataset.

Conclusions:

• YATSIDroid offers a highly effective solution for Android malware detection.
• The proposed model excels in scenarios with limited labeled data.
• This approach significantly enhances the security of smartphone applications against evolving threats.