ウダ　リュウヤ   宇田　隆哉    所属   コンピュータサイエンス学部 コンピュータサイエンス学科    職種   准教授
言語種別	英語
発行・発表の年月	2025/02
形態種別	国際会議論文
査読	査読あり
標題	Machine Learning Approach to Malware Classification Using Byte n-grams on IoT Devices
執筆形態	共著
掲載誌名	Proceedings of the 2025 19th International Conference on Ubiquitous Information Management and Communication
掲載区分	国外
出版社・発行元	IEEE
総ページ数	6
担当区分	最終著者
著者・共著者	Yuto Yama and Ryuya Uda
概要	Malware for IoT devices has become popular in recent years. Many detection methods have been proposed to detect the malware, mainly using machine learning, but they are not designed to work on IoT devices and are often implemented using Python. Running a Python program requires the installation of a package, which is impractical given the memory and storage size of IoT devices. In addition, methods that require decompilation, static and dynamic analysis of test samples are difficult to run on IoT devices. Therefore, in this research, we propose a method implemented in C/C++ so that malware detection using byte n-grams and machine learning can be run on IoT devices, and then resource consumption can be measured. Byte n-gram methods are good approaches to detect malware without knowing the execution environment. Moreover, the Top-L approach to information gain can be effectively applied to reduce storage and memory consumption. As a result of the evaluation, when using an SVM model with Top-50000 that maintains effective classification accuracy, the memory consumption was 71.91MB and the storage consumption was 1.3GB, which can be implemented in IoT devices.