Machine Learning based Hybrid Approach for Prediction Model of Android Malware

Abstract

In this paper, a hybrid machine learning based model (MLP-DT) is proposed to achieve an effective approach in order to increase the detection accuracy and reduce the computational complexity in predicting Android malware. The suggested model combines MLP with the power of decision making in DT, which will enable it to classify Malware and Benign Android apps efficiently. Android application datasets are extracted for features that are relevant to the extraction process (permissions, api use, behavioral patterns), and these are preprocessed, and then normalized. The MLP model is used to identify non-linear relationships in the data, and the Decision-Tree model is used to further enhance the classification process, increase interpretability, and reduce false positives. The hybrid framework combines the advantages of both techniques and achieves better predictions as compared to the individual techniques. The proposed approach has been experimentally evaluated and shown to yield better accuracy, precision and recall which in turn can be used to detect the malware in cyber security applications for real time Android platforms.