Parameter Optimization of OCSVM by GridSearchCV for Improving IoT Malware Detection Accuracy

is written in 200-300 words. In recent years, the Internet of Things (IoT) has been playing an increasingly important role in our lives. It has enabled us to form new services and business models, and to automate and improve the efficiency of our work. However, at the same time, security vulnerabilities have become an issue: according to NICT, more than half of the observed cyber attack-related communications targeted the IoT [3].In this study, we propose a system that combines n-gram analysis and OCSVM to discriminate between normal communication of IoT devices and communication after malware infection, and a system that uses GridSearchCV to calculate the best parameters for the parameters arbitrarily set by humans when applying OCSVM. The system is proposed to calculate the best parameters using GridSearchCV. For the evaluation, we conducted an experiment to compare the detection accuracy of 25 sets of parameters conventionally used and the detection accuracy using GridSearchCV, and showed good detection accuracy. However, in some cases, the detection accuracy decreased in the process of increasing the total number of data given to GridSearchCV. As a future subject, it is necessary to change the ratio and the total number of normal data and more than normal data, to observe the change of detection accuracy, and to study the cause.