Related Experiment Video
Updated: Jul 13, 2026
07:35
Selecting Multiple Biomarker Subsets with Similarly Effective Binary Classification Performances
Published on: October 11, 2018
7.5K
PermDroid a framework developed using proposed feature selection approach and machine learning techniques for Android
Arvind Mahindru1, Himani Arora2, Abhinav Kumar3
1Department of Computer Science and applications, D.A.V. University, Sarmastpur, Jalandhar, 144012, India. er.arvindmahindru@gmail.com.
Scientific Reports
|May 10, 2024
Summary
This study introduces a novel feature selection framework to improve Android malware detection. The proposed method enhances machine learning model accuracy, achieving 98.8% detection for Android malware.
Area of Science:
- Computer Science
- Cybersecurity
- Machine Learning
Background:
- Android malware detection is challenging due to its permission model.
- Previous methods often used excessive features, leading to overburdened models.
- Effective machine learning relies on relevant, discriminative features.
Purpose of the Study:
- To propose a feature selection framework for Android malware detection.
- To identify relevant features that enhance model accuracy and reduce misclassification.
- To develop and evaluate machine learning models using selected features.
Main Methods:
- A two-stage feature selection framework was implemented.
- Stage 1: t-test and univariate logistic regression.
- Stage 2: Multivariate linear regression and correlation analysis; models built with ensemble methods and neural networks.
Main Results:
- The feature selection framework identified relevant features for malware detection.
- Models using selected features outperformed those using all extracted features.
- The developed model achieved a high accuracy of 98.8% on half a million Android apps.
Conclusions:
- The proposed feature selection framework effectively enhances Android malware detection.
- Optimized feature sets lead to more accurate and efficient machine learning models.
- This approach offers a significant improvement over existing Android malware detection methodologies.
