Android Malware Detection Using Multi-domain Feature Analysis And Deep Learning Models

Abstract

We Propose A Hybrid And Intelligent Android Malware Detection Framework That Integrates Multi-domain Feature Extraction With Deep Learning Models To Improve Mobile Application Security. The System Consists Of Two Key Modules: The APK Security Analyzer And The App Behavior Analyzer. The APK Security Analyzer Applies Both Static And Dynamic Analysis Using Tools Such As Androguard, Android Emulator, ADB, And Monkey To Extract Permissions, API Call Sequences, And Network Traffic Features. These Are Processed Using A Tri-model Architecture—Multi-Layer Perceptron (MLP), Long Short-Term Memory (LSTM), And LightGBM—within A Streamlitinterface.The App Behavior Analyzer Enhances Static CSV-based Analysis By Automating Preprocessing With StandardScaler And Applying A Sliding Window For Time-series Classification Using An LSTM Model. It Supports Real-time Predictions And Provides Interactive Visualizations To Aid User Understanding Of Results.Experimental Results Show High Accuracy: MLP And LSTM Models Each Achieved 96%, While LightGBM Reached 88%. Precision, Recall, And F1-scores Confirm System Robustness, With MLP Scoring 96%, 93%, And 94%, Respectively. Visualizations Such As Radar Charts And Progress Bars Clearly Communicate App Risk Levels And Behaviorpatterns.These Findings Establish The Framework As A Robust, Scalable, And Interpretable Solution For Real-time Android Malware Detection, Advancing Mobile Cybersecurity Tools.

Keywords