Comparisons of machine learning techniques for detecting malicious webpages

作者：

Highlights：

• 3 supervised and 2 unsupervised techniques are modeled to detect malicious webpages.

• Supervised machine learning (ML) techniques accuracies are above 89%.

• Unsupervised ML techniques accuracies have at least a silhouette coefficient of 0.87.

• Information obtained from URLs, page links, semantics and visual features of webpages.

• Chrome extension, lightweight and heavyweight classifiers and online learning are used.

摘要

•3 supervised and 2 unsupervised techniques are modeled to detect malicious webpages.•Supervised machine learning (ML) techniques accuracies are above 89%.•Unsupervised ML techniques accuracies have at least a silhouette coefficient of 0.87.•Information obtained from URLs, page links, semantics and visual features of webpages.•Chrome extension, lightweight and heavyweight classifiers and online learning are used.

论文关键词：K-Nearest Neighbor,Support Vector Machine,Naive Bayes,Affinity Propagation,K-Means,Supervised and unsupervised learning

论文评审过程：Available online 16 September 2014.

论文官网地址：https://doi.org/10.1016/j.eswa.2014.08.046