Enhancement of Machine Learning Models Using Intersection Filtering Model Based on Recursive Feature Elimination on Specified Android Malware Classification

The rapid growth of Android devices has made them a primary target for malware attacks, posing significant security risks.Traditional malware detection methods are increasingly ineffective against sophisticated and obfuscated malware, driving the need for machine learning (ML) models capable of early malware detection.However, the high dimensionality of Android malware datasets, filled with irrelevant and redundant features, often degrades ML performance.To address these challenges, this paper introduces an Intersection Filtering Model based on Recursive Feature Elimination (RFE) to enhance ML models in Android malware detection.The proposed model employs an intersection filtering mechanism that combines feature sets generated from multiple selection methods, retaining only features consistently deemed important.This refined feature set is then processed by RFE to further eliminate redundant and noisy features.The model significantly improves classification accuracy, precision, recall, and computational efficiency by streamlining feature selection.Experimental results demonstrate that the Intersection Filtering Model outperforms traditional feature selection methods, offering a robust solution for early Android malware detection.The resulting performance of the proposed method is achieved the effectiveness level more than 2%-3% in each scenario than that of using RFE alone or the original set in all tests on the three Android Malware Classifications (TuandroMD, Drebin and CIC-2020 Android Malware).This proposed method can also be implemented in other high-dimensional machine learning tasks, providing a basis for improving performance in various malware detection datasets on Android devices.