WEKO3
アイテム
侵入検知システムの検知精度向上のための敵対的サンプルによるデータ拡張法の検討
http://hdl.handle.net/10458/00010278
http://hdl.handle.net/10458/0001027872767650-71a6-4690-a36a-e88f1e3ffb79
| 名前 / ファイル | ライセンス | アクション |
|---|---|---|
本文 (1.8 MB)
|
|
| Item type | 紀要論文 / Departmental Bulletin Paper(1) | |||||
|---|---|---|---|---|---|---|
| 公開日 | 2021-10-26 | |||||
| タイトル | ||||||
| タイトル | 侵入検知システムの検知精度向上のための敵対的サンプルによるデータ拡張法の検討 | |||||
| 言語 | ja | |||||
| タイトル | ||||||
| タイトル | Data Expansion Using Adversarial Examples to Improve the Accuracy of Intrusion Detection Systems | |||||
| 言語 | en | |||||
| 言語 | ||||||
| 言語 | jpn | |||||
| キーワード | ||||||
| 言語 | en | |||||
| 主題Scheme | Other | |||||
| 主題 | Intrusion detection system | |||||
| キーワード | ||||||
| 言語 | en | |||||
| 主題Scheme | Other | |||||
| 主題 | Data expansion | |||||
| キーワード | ||||||
| 言語 | en | |||||
| 主題Scheme | Other | |||||
| 主題 | Adversarial example | |||||
| キーワード | ||||||
| 言語 | en | |||||
| 主題Scheme | Other | |||||
| 主題 | Jacobian-based saliency map attack | |||||
| 資源タイプ | ||||||
| 資源タイプ識別子 | http://purl.org/coar/resource_type/c_6501 | |||||
| 資源タイプ | departmental bulletin paper | |||||
| 著者 |
川畑, 魁星
× 川畑, 魁星× 金丸, 和樹× 油田, 健太郎× 山場, 久昭× 岡崎, 直宣× Kawabata, Kaisei× 金丸, 和樹 |
|||||
| 抄録 | ||||||
| 内容記述タイプ | Abstract | |||||
| 内容記述 | In recent years, cyber-attacks such as unauthorized access and malware have been increasing along with the increasing use of network systems and the spread of new network technologies. Intrusion detection systems (IDS) have been attracting attention as one of the security technologies to protect systems from these cyber attacks. The accuracy of intrusion detection using deep learning is highly dependent on the data used for training, and a large amount of labeled training data is required. It is difficult to prepare a large amount of training data while taking into account the bias of the data. In this research, we propose a computationally efficient data expansion method using Jacobian-based Saliency Map Attack (JSMA), one of the adversarial example generation methods, and investigate how to improve the detection accuracy of signature-based IDS using deep learning models. To evaluate the proposed method, we built a small-scale model, extracted data with low classification accuracy, and extended the data with adversarial samples that were perturbed to bring them closer to the correct class, and compared the detection accuracy before and after the data extension. As a result of the experiment, the detection accuracy after the data expansion using the adversarial example was found to be better than that before the data expansion in terms of Accuracy, Recall, and F1-score. Although the proposed method improves the detection performance against attacks, it also increases the possibility of false positives, which requires improvements to reduce the degradation of Precision. | |||||
| 言語 | en | |||||
| 書誌情報 |
ja : 宮崎大学工学部紀要 en : Memoirs of Faculty of Engineering, University of Miyazaki 巻 50, p. 125-129, 発行日 2021-09-28 |
|||||
| 出版者 | ||||||
| 出版者 | 宮崎大学工学部 | |||||
| 言語 | ja | |||||
| 出版者 | ||||||
| 出版者 | Faculty of Engineering, University of Miyazaki | |||||
| 言語 | en | |||||
| ISSN | ||||||
| 収録物識別子タイプ | ISSN | |||||
| 収録物識別子 | 05404924 | |||||
| 書誌レコードID | ||||||
| 収録物識別子タイプ | NCID | |||||
| 収録物識別子 | AA00732558 | |||||
| 著者版フラグ | ||||||
| 出版タイプ | VoR | |||||
| 出版タイプResource | http://purl.org/coar/version/c_970fb48d4fbd8a85 | |||||
