Advanced Risk Analyzer for Android Apps

Abstract

The rapid growth of Android applications has changed mobile technology by making it accessible, innovative, and convenient for billions of users around the world. However, this popularity has also led to an increase in malicious apps that threaten user privacy, exploit sensitive data, and pose serious security risks. Traditional detection methods, like signature- based antivirus solutions, often fail against zero-day malware, obfuscated code, or overly permissive applications that seem safe but misuse data. To tackle these issues, we present AndroidRiskCheck, a hybrid framework that combines static program analysis with machine learning (ML) to evaluate risks in Android applications. The system extracts permissions and metadata directly from APK files using a Python-based pipeline, removing Java dependencies and ensuring light cross-platform performance. A smart XML parsing mechanism is used to manage both human-readable and binary AndroidManifest files, ensuring reliability across different APK structures. The features extracted are converted into numerical vectors and classified with an ensemble of three ML models: Multinomial Naive Bayes, Logistic Regression, and Gradient Boosting. This categorizes applications as Safe, Risky, or Malicious. A risk scoring system, permission categorization, and compliance checks with GDPR, CCPA, and DPDP regula- tions further improve the system’s clarity and practical use. The tool includes a Flask-based web interface to provide an easy upload-and-analysis process, along with visualization dashboards for risk scores and tracker detection. Additionally, the system allows optional extensions such as dynamic analysis (using Frida and mitmproxy), threat intelligence integration (like the VirusTotal API), and blockchain-based report storage for auditability and integrity. By combining lightweight static analysis, strong ML-driven classification, and privacy compliance checks, AndroidRiskCheck offers an effective, accessible, and adaptable solution for evalu- ating Android security.