An Enhanced Machine Learning–Based Android Malware Detection Framework with Static Analysis

Abstract

The widespread use of Android devices has made them a prime target for increasingly sophisticated malware attacks, posing serious threats to user privacy and data security. Traditional signature-based detection methods fail to identify novel and polymorphic malware variants, necessitating more adaptive approaches. This paper presents an improved machine learning-based framework for Android malware detection that overcomes the limitations of existing systems through static analysis techniques. Our framework leverages advanced feature engineering methods to extract comprehensive behavioral and structural characteristics from Android applications, such as permissions, API calls, network activities, and code-level static attributes. We implement and evaluate several machine learning algorithms to achieve robust classification performance. The proposed framework uses feature selection optimization to reduce dimensionality while maintaining detection accuracy, balancing computational efficiency with detection effectiveness. Experimental evaluation on benchmark dataset demonstrates that our framework achieves good performance compared to state-of-the-art approaches, ensuring the detection of zero-day malware variants. The results indicate that our enhanced framework offers a practical and scalable solution for real-time Android malware detection in various deployment scenarios.