Published Paper Details:

With the explosive growth of Android applications, mobile malware poses an ever increasing threat to user privacy and device security. Traditional signature based detectors struggle against obfuscated or zero day malware, necessitating intelligent, data driven solutions. This paper presents a lightweight, hybrid Android malware detection framework that combines static feature extraction with an Equilibrium Optimizer (EO)based feature selection module to reduce dimensionality and highlight the most informative attributes. A hybrid ensemble of LightGBM, XGBoost, Random Forest, and a Bidirectional LSTM (Bi-LSTM) model is then employed to classify applications as benign or malicious. The entire pipeline is exposed via a Flask-based REST API, supporting real-time APK uploads, JSON outputs, and SQLite-backed logging. Experimental evaluation on a dataset of 12,000+ APKs achieves an overall accuracy of 95.2%, precision of 94.7%, recall of 95.8%, and F1 score of 95.2%, significantly outperforming baseline methods. The proposed system demonstrates robust detection capabilities, low computational overhead, and easy deploy ability for proactive Android security.