Lightweight On-Device Detection of Android Malware Based on the Koodous Platform and Machine Learning

Area of Science:

• Cybersecurity
• Machine Learning
• Mobile Security

Background:

• Android's dominance in mobile OS increases vulnerability to malware.
• Smartphones are critical for sensitive activities like e-banking and e-identity verification.
• Existing security tools like BotSense Mobile protect critical applications.

Purpose of the Study:

• To introduce novel malware detection functionality for BotSense Mobile.
• To develop and evaluate a machine learning model for identifying unknown malicious Android applications.
• To ensure user data privacy by deploying the model on edge devices.

Main Methods:

• Developed a lightweight neural network for malware detection.
• Deployed the model on edge devices, utilizing only manifest-related features.
• Conducted empirical analysis using recent data from the Koodous platform (May-June 2022).

Main Results:

• The machine learning model achieved an f1-score of 0.77 and a precision of 0.9 on recent data.
• Highlighted the challenge of machine learning model aging in malware detection.
• Demonstrated the feasibility of on-device malware detection using lightweight models.

Conclusions:

• The proposed machine learning model offers effective detection of unknown Android malware.
• Edge deployment and manifest-feature utilization maintain user data privacy.
• Addressing model aging is crucial for sustained performance in mobile malware detection systems.