Detecting Malicious Websites Based on JavaScript Content Analysis

• Item type: Symposium(1)
• 公開日: 2021-10-19
• タイトル: Detecting Malicious Websites Based on JavaScript Content Analysis
• タイトル / 言語: en
• タイトル: Detecting Malicious Websites Based on JavaScript Content Analysis
• 言語: eng
• キーワード / 主題Scheme: Other
• キーワード / 主題: Web-based Attack，Machine Learning，JavaScript，graph2vec，representation learning
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_5794
• 資源タイプ: conference paper
• 著者所属: 情報通信研究機構／神戸大学
• 著者所属: 情報通信研究機構
• 著者所属: 神戸大学
• 著者所属: 神戸大学
• 著者所属: 情報通信研究機構
• 著者所属: 情報通信研究機構
• 著者所属(英): en
• 著者所属(英): National Institute of Information and Communications Technology / Graduate School of Engineering, Kobe University
• 著者所属(英): en
• 著者所属(英): National Institute of Information and Communications Technology
• 著者所属(英): en
• 著者所属(英): Graduate School of Engineering, Kobe University
• 著者所属(英): en
• 著者所属(英): Graduate School of Engineering, Kobe University / Center for Mathematical and Data Sciences, Kobe University
• 著者所属(英): en
• 著者所属(英): National Institute of Information and Communications Technology
• 著者所属(英): en
• 著者所属(英): National Institute of Information and Communications Technology
• 著者名: Muhammad, Fakhrur Rozi; 班, 涛; Sangwook, Kim; 小澤, 誠一; 高橋, 健志; 井上, 大介
• 著者名(英): Muhammad, Fakhrur Rozi; Tao, Ban; Sangwook, Kim; Seiichi, Ozawa; Takeshi, Takahashi; Daisuke, Inoue
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: Analyzing JavaScript contents has been a promising way to detect malicious websites. However, attackers often put malicious scripts in unreachable places and use complicated obfuscation tools to hinder the detection. Therefore, finding malicious scripts exhaustively inside the website tend to be time-consuming and ineffective in improving detection performance. To address these challenges, we introduce a novel approach to detecting malicious websites by analyzing the collective representation of the stack of JavaScripts in a website. First, we build a collective graph representation of a website by aggregating abstract syntax trees of all JavaScripts therein. Then, we use graph2vec to encode the graph into a vectorial representation. Finally, machine learning based detection is performed for identifying potentially harmful websites. Results showed that the proposed approach achieves high accuracy on a real-world dataset, outperforming prior approaches. We believe the result in this paper can open new opportunities for more effective malicious-website detection.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: Analyzing JavaScript contents has been a promising way to detect malicious websites. However, attackers often put malicious scripts in unreachable places and use complicated obfuscation tools to hinder the detection. Therefore, finding malicious scripts exhaustively inside the website tend to be time-consuming and ineffective in improving detection performance. To address these challenges, we introduce a novel approach to detecting malicious websites by analyzing the collective representation of the stack of JavaScripts in a website. First, we build a collective graph representation of a website by aggregating abstract syntax trees of all JavaScripts therein. Then, we use graph2vec to encode the graph into a vectorial representation. Finally, machine learning based detection is performed for identifying potentially harmful websites. Results showed that the proposed approach achieves high accuracy on a real-world dataset, outperforming prior approaches. We believe the result in this paper can open new opportunities for more effective malicious-website detection.
• 書誌情報: コンピュータセキュリティシンポジウム2021論文集; p. 727-732, 発行日 2021-10-19
• 出版者 / 言語: ja
• 出版者: 情報処理学会