Poster: Android Malware Detection using Hybrid Features and Machine Learning

The proliferation of the Android platform has in-creased malware development. New or zero-day attacks easily bypass traditional anti-malware techniques relying on signature-based detection. Recent machine learning (ML) models for An-droid malware detection struggle with obfuscated malware and high false-positive rates. This paper proposes a new technique for detecting Android malware using static hybrid features and ML algorithms. The proposed model enhances training efficiency and generalization by removing weak features during the feature selection. We use various ML algorithms like Logistic Regression (LR) and Random Forest (RF) to train the detection model on the OmniDroid dataset, which includes 22,000 malware and benign samples. Experimental results show that our model achieved 90% accuracy and 91% precision.