Ads-Portal Domains: Identification and Measurements

作者：Almishari Mishari^*; Yang Xiaowei

来源：ACM Transactions on the Web, 2010, 4(2): 4.

DOI：10.1145/1734200.1734201

• Measurement
• Security
• Algorithms
• Experimentation
• Ads-portal
• parking service
• parked domain
• advertisement syndication
• data mining
• Web characterization

An ads-portal domain refers to a Web domain that shows only advertisements, served by a third-party advertisement syndication service, in the form of ads listing. We develop a machine-learning-based classifier to identify ads-portal domains, which has 96% accuracy. We use this classifier to measure the prevalence of ads-portal domains on the Internet. Surprisingly, 28.3/25% of the (two-level) *.com. net web domains are ads-portal domains. Also, 41/39.8% of *.com. net ads-portal domains are typos of well-known domains, also known as typo-squatting domains. In addition, we use the classifier along with DNS trace files to estimate how often Internet users visit ads-portal domains. It turns out that similar to 5% of the two-level *.com, *.net, *.org, *.biz and *.info web domains on the traces are ads-portal domains and similar to 50% of these accessed ads-portal domains are typos. These numbers show that ads-portal domains and typo-squatting ads-portal domains are prevalent on the Internet and successful in attracting many visits. Our classifier represents a step towards better categorizing the web documents. It can also be helpful to search engines ranking algorithms, helpful in identifying web spams that redirects to ads-portal domains, and used to discourage access to typo-squatting ads-portal domains.

• 出版日期2010-4