Machine learning classification model for Network based Intrusion Detection System

Evaluation of ensemble machine learning methods in mobile threat detection

“…The dataset was used to build the ML Classifiers using WEKA [30]. In this study, the focus is only on the evaluation of ensemble ML classification model, as the evaluation of individual classifiers is already done in our previous work [10].…”

“…1, comprised of Traffic generation, preprocessing, model building and evaluation of the ML model. Traffic was generated for both benign and malicious applications using the method mentioned in our previous research [10]. A number of benign applications were used to generate real traffic which were installed from Google playstore.…”

“…Most of the research done in the field of using ML techniques to detect Android-based malware is based on features such as system or API calls. There are only a few studies which have focused on network-based intrusion detection for android malwares [10] [11]. In some of the studies [12] [7] [13] [14], the detection engine or model need to be installed on the mobile phone to detect e.g.…”

“…The performance evaluation of these algorithms was already performed individually in our previous work [10]. In this study, we have used ensemble methods to combine the output of these ML algorithms to increase the effectiveness of the ML classifiers.The combination methods used in our study such as majority voting, maximum probability and product of probabilities were adopted from [21] [22].…”

“…These datasets are quite old and not applicable to mobile attacks. Due to non-availability of public datasets in this domain, we have used a dataset which was created in our previous research [10] for evaluation purpose. This dataset was build using the traffic generated by several benign and malicious samples.…”