Efficient Machine Learning Method for Protocol Fuzz Testing: Improvement of Sequence-to-Sequence Model and Refined Training Data

• Item type: SIG Technical Reports(1)
• 公開日: 2022-03-04
• タイトル: Efficient Machine Learning Method for Protocol Fuzz Testing: Improvement of Sequence-to-Sequence Model and Refined Training Data
• タイトル / 言語: en
• タイトル: Efficient Machine Learning Method for Protocol Fuzz Testing: Improvement of Sequence-to-Sequence Model and Refined Training Data
• 言語: eng
• キーワード / 主題Scheme: Other
• キーワード / 主題: テスト，運用・保守
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_18gh
• 資源タイプ: technical report
• 著者所属: JVCKENWOOD Corporation
• 著者所属: JVCKENWOOD Corporation
• 著者所属: JVCKENWOOD Corporation
• 著者所属: Nagoya University
• 著者所属(英): en
• 著者所属(英): JVCKENWOOD Corporation
• 著者所属(英): en
• 著者所属(英): JVCKENWOOD Corporation
• 著者所属(英): en
• 著者所属(英): JVCKENWOOD Corporation
• 著者所属(英): en
• 著者所属(英): Nagoya University
• 著者名: Bo, Wang; Toshihiro, Maruyama; Ako, Suzuki; Yuichi, Kaji
• 著者名(英): Bo, Wang; Toshihiro, Maruyama; Ako, Suzuki; Yuichi, Kaji
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: Fuzz testing is one of software testing methods for finding software vulnerabilities and is used as a technology for finding unknown security vulnerabilities as a black-box test. Although many fuzz testing methods that are based on machine learning have been investigated, they cannot analyze and learn the real-time status of communication protocol. We focus on the method of efficient machine learning for protocol fuzzing, and present major problems of current fuzzing tools, and introduce techniques to get around the problems with an improvement of Sequence-to-Sequence model and refined training.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: Fuzz testing is one of software testing methods for finding software vulnerabilities and is used as a technology for finding unknown security vulnerabilities as a black-box test. Although many fuzz testing methods that are based on machine learning have been investigated, they cannot analyze and learn the real-time status of communication protocol. We focus on the method of efficient machine learning for protocol fuzzing, and present major problems of current fuzzing tools, and introduce techniques to get around the problems with an improvement of Sequence-to-Sequence model and refined training.
• 書誌レコードID / 収録物識別子タイプ: NCID
• 書誌レコードID / 収録物識別子: AN10112981
• 書誌情報: 研究報告ソフトウェア工学（SE）; 巻 2022-SE-210, 号 32, p. 1-7, 発行日 2022-03-04
• ISSN / 収録物識別子タイプ: ISSN
• ISSN / 収録物識別子: 2188-8825
• Notice: SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc.
• 出版者 / 言語: ja
• 出版者: 情報処理学会