An automatic malware detection system for Android Comparative analysis for Android malware detection Android malware detection using network traffic analysis Mobile devices, which are becoming more and more widespread today, have turned into hand-held computers thanks to the multimedia communication and applications. Today, the multimedia applications have been supported by traditional mobile phones. Users can use their mobile devices for many purposes such as internet access, online banking, social networks, file sharing and entertainment. The ability to perform transactions such as financial transactions, online shopping and sensitive data transfers on mobile devices with increased functionality makes mobile devices the target of attackers. In this study, a deep learning based malware detection system has been developed based on the interactions of mobile applications on the network. The developed LSTM-based deep learning model has been analyzed comparatively with NB, RF, SVM, MLP and CNN using accuracy, precision, recall and F-1 metrics. The experimental results showed that the developed LSTM-based deep learning model is more successful in malware detection than others with 95% accuracy.Figure A. Proposed LSTM model for Android malware classification Purpose: In this study, an automatic detection system for Android malware is proposed. To illustrate the results of the proposed LSTM based model, a comprehensive and comparative experimental study has been carried out.
Theory and Methods:With the developed LSTM-based deep learning model, it is aimed to detect Android malware by using network traffic data. In addition, the proposed model compared with the existing deep learning algorithms (i.e., MLP and CNN) and conventional machine learning algorithms (i.e., NB, RF and SVM).
Results:In experimental studies, the LSTM has shown a more successful classification performance in detecting malware compared to other models.
Conclusion:Experimental results showed that the accuracy values of LSTM, CNN, ML SVM, RF and NB are 0.950, 0.911, 0.738, 0.562, 0.912 and 0.446, respectively. The developed LSTM-based model achieved more successful results for each of the accuracy, precision and recall and f-1 metrics compared to other models.