FSDroid:- A feature selection technique to detect malware from Android using Machine Learning Techniques: FSDroid.

Area of Science:

• Computer Science
• Cybersecurity
• Machine Learning

Background:

• Android's widespread adoption has led to an increase in malware-infected applications.
• The Android permission model presents vulnerabilities exploited by cybercriminals to steal user data.
• Detecting and removing malware from app stores is a critical challenge.

Purpose of the Study:

• To develop an effective malware detection system for Android applications.
• To investigate the impact of app permissions on security risks.
• To identify optimal features for malware detection using various selection approaches.

Main Methods:

• Implemented ten distinct feature selection methods to identify key indicators of malware.
• Developed a malware detection model using Least Square Support Vector Machine (LSSVM) with linear, radial basis, and polynomial kernels.
• Trained and evaluated the model on a dataset of 200,000 Android applications.

Main Results:

• The FS-droid model, utilizing LSSVM with a radial basis function kernel, achieved a malware detection rate of 98.8%.
• FS-droid demonstrated a 3% higher detection rate compared to existing anti-virus scanners.
• The system outperformed other proposed frameworks and approaches in malware detection accuracy.

Conclusions:

• The proposed FS-droid system offers a highly effective solution for detecting Android malware.
• Analyzing app permissions and employing advanced machine learning techniques are crucial for enhancing mobile security.
• The study highlights the potential of LSSVM with RBF kernel for robust malware identification.