V2I11P15

Survey on Machine Learning-Based Android Adware Detection

Mane Pranali Rajendra^1*, Dr. Sanjaykumar P. Pingat²

Abstract

The Android operating system uses a permission-based security mechanism to limit application access to device resources. However, this approach is frequently inefficient since people provide rights without careful consideration, allowing hostile entities such as malware to abuse device features. This study describes a machine learning-based method for detecting Android adware by evaluating permission requests, intents, and API usage patterns. The system is trained and validated using publicly available datasets and multiple machine learning classifiers. Key characteristics are pre-processed, including normalization and dimensionality reduction, to improve classification accuracy. The experimental results demonstrate that combining machine learning with static app metadata effectively detects adware, indicating the potential of lightweight and scalable security solutions for Android smartphones.

Keywords:

Malware Detection, Feature Extraction, Machine Learning, Support Vector Machine