Android Malware Multiclass Classification using Machine Learning: Evaluating the Performance of Random Forest, Artificial Neural Network, and Convolutional Neural Network
Journal of Logistics Informatics and Service Science · 2024
This research investigates the effectiveness of machine learning techniques, namely random forest, artificial neural network, and convolutional neural network, in detecting and classifying Android malware using both static and dynamic analysis methods.Leveraging the CICInvesAndMal2019 dataset, multiclass classification by categorizing malware into adware, ransomware, scareware, and SMS malware are performed.The static analysis examines permissions and intents, while the dynamic analysis focuses on API calls and network flows.The performance of the models is evaluated using accuracy, recall, precision, F1 score, training time, and testing time.The results demonstrate the superiority of random forest over deep learning models in both static and dynamic analysis, with static analysis yielding better performance than dynamic analysis.This study contributes to the field of Android malware detection by providing insights into the effectiveness of different machine learning algorithms and analysis methods, highlighting the potential of random forest for efficient and accurate malware multiclass classification.