講演抄録/キーワード |
講演名 |
2020-10-01 11:15
Malicious URLs Detection Using an Integrated AI Framework ○Bo-Xiang Wang・Ren-Feng Deng・Yi-Wei Ma・Jiann-Liang Chen(NTUST) IA2020-1 |
抄録 |
(和) |
Malicious attacks on computer networks are quite common, and the internet attacks are even more widespread, such as Malvertising, Phishing, and Drive-by download, all of which are related to malicious URL links. The conventional way to prevent these malicious URLs would be to manage them through a blacklist, that requires considerable human resources to identify them. In recent years, with the improvement of hardware and software devices, computers with machine learning are able to learn and predict from large amounts of data, therefor replacing traditional methods and saving manpower. This study proposed an integrated AI framework, which consists of a fast filtering component and a precise identification component. This framework combines the advantages of the CNN (Convolutional Neural Network) model and the XGBoost (eXtreme Gradient Boosting) model to achieve a fast and accurate detection capability. Experimental results show that the fast filter is able to detect results in 0.6 seconds with an accuracy of 83%. In contrast, the accuracy of the precision identification component is 94% when it takes about 40 seconds to detect the result. This study integrates the advantages of the two components to achieve the goal of fast and accurate malicious URL detection. |
(英) |
Malicious attacks on computer networks are quite common, and the internet attacks are even more widespread, such as Malvertising, Phishing, and Drive-by download, all of which are related to malicious URL links. The conventional way to prevent these malicious URLs would be to manage them through a blacklist, that requires considerable human resources to identify them. In recent years, with the improvement of hardware and software devices, computers with machine learning are able to learn and predict from large amounts of data, therefor replacing traditional methods and saving manpower. This study proposed an integrated AI framework, which consists of a fast filtering component and a precise identification component. This framework combines the advantages of the CNN (Convolutional Neural Network) model and the XGBoost (eXtreme Gradient Boosting) model to achieve a fast and accurate detection capability. Experimental results show that the fast filter is able to detect results in 0.6 seconds with an accuracy of 83%. In contrast, the accuracy of the precision identification component is 94% when it takes about 40 seconds to detect the result. This study integrates the advantages of the two components to achieve the goal of fast and accurate malicious URL detection. |
キーワード |
(和) |
Malicious URL / Integrated AI framework / Artificial Intelligence / Feature Selection / / / / |
(英) |
Malicious URL / Integrated AI framework / Artificial Intelligence / Feature Selection / / / / |
文献情報 |
信学技報, vol. 120, no. 177, IA2020-1, pp. 1-5, 2020年10月. |
資料番号 |
IA2020-1 |
発行日 |
2020-09-24 (IA) |
ISSN |
Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
IA2020-1 |
|