Classifying android malware to secure mobile platforms using machine learning and deep learning approaches

The widespread use of Android devices has made them a prime target for cyberattacks, leading to an alarming rise in mobile malware. Detecting and classifying such threats effectively is essential to securing user data and maintaining platform integrity. This study presents a comprehensive framework for Android malware classification using a combination of machine learning (ML), deep learning (DL), and hybrid ensemble approaches. The research utilizes the CICMalDroid2020 dataset, containing 11,598 Android applications categorized into five classes: Adware, Banking Malware, SMS Malware, Riskware, and Benign. The data underwent extensive preprocessing, including normalization, label encoding, balancing, and feature extraction using the LASSO technique. Various ML algorithms such as Random Forest (RF), Decision Tree (DT), K-Nearest Neighbor (KNN), Support Vector Machine (SVM), and Extreme Gradient Boosting (XGB) were implemented alongside DL models like ANN, CNN, RNN, and LSTM. Additionally, hybrid models combining CNN with LSTM, XGB, and RF were developed to enhance detection accuracy. Experimental results indicate that the proposed models outperform existing approaches in malware classification. The XGB model achieved the highest accuracy of 95.74%, followed by the Ensemble (RF, ET, XGB) with 95.44%, surpassing previous work. These results demonstrate that integrating ensemble and deep learning architectures provides improved generalization, robustness, and precision in detecting Android malware.