<![CDATA[This study aims to develop an efficient Android malware detection framework using the LightGBM algorithm optimized through Chi-Square feature selection and hyperparameter tuning. Although machine learning techniques have been widely applied in Android malware detection, many previous studies primarily focus on classification accuracy while paying limited attention to feature reduction efficiency, computational complexity, and permission-based feature interpretability. Therefore, this study proposes a lightweight permission-based detection framework that integrates Chi-Square feature selection with optimized LightGBM classification to improve detection performance while reducing irrelevant features. The dataset used was obtained from Kaggle and consists of 29,300 Android applications, including 14,630 benign applications and 14,700 malware samples. Each application is represented using 86 binary permission-based features extracted from the AndroidManifest.xml file. The research stages include preprocessing, feature selection, training-testing data splitting, hyperparameter optimization, and model evaluation. Experimental results show that the proposed model achieves an accuracy of 0.96, precision of 0.95, recall of 0.96, F1-score of 0.96, and ROC-AUC of 0.9885. These findings indicate that the proposed framework effectively distinguishes malware from benign applications while maintaining computational efficiency and improving permission-based malware detection interpretability. The novelty of this study lies in the integration of Chi-Square feature selection and Grid Search-optimized LightGBM to develop a lightweight permission-based Android malware detection framework that reduces feature dimensionality while maintaining high detection performance and interpretability.]]>
Copyrights © 2026
