Classification of concurrent anomalies for iot software based support vector machine

Abstract

Internet of Thing (IoT) can connect anyone with anything at any point in any place. Currently, growing number of IoT devices have become a major role of daily life owing to their convenience. The IoT devices usually controlled by Web applications and mobile applications, which will process lots of events from user’s controller to devices. Hence, such software is a kind of concurrent program in IoT environment because the software is unable to simultaneously process these events, which may cause the concurrent issue. There is event-drive model in either Web application or mobile applications, which is unable to easily detect the concurrent anomaly by existing approaches due to the non-determined of execution and hardly reproduced by the same sequence. The previous techniques of concurrent detection are excessive limitations that only used for one of concurrent anomaly with the large number of false positive. In this paper, we describe a novel methodology to dynamically classify two types of concurrent anomalies for IoT software. According to the executable sequence graph, we generate the training and test examples for classification. The vectorization features are classified by Support Vector Machine (SVM) with Gaussian kernel. The SVM will predict the concurrent state of current executable example. As a result, the optimal true positive of simulation is 80% in our experiment which is a higher accuracy than others.