Improvement of Referrer SPAM blocking system

Abstract

In recent years, the use of Weblog is increasing rapidly and some people are using functions to record referer (URL of the page that the site visitor was viewing immediately before) and create back-links to the URL. Recently, as the main purpose to guide visitors to harmful sites, people are using this function to misinterpret referer information and access a large number of sites indiscriminately causing the problem called referrer spam. To combat the referrer spam, the authors have proposed the referrer spam blocking system using Bayesian filter. However, the proposed system has non-negligible false negative (i.e. undetected spam request) ratio of 12%. In this paper, we improve the previous proposed system by introducing image optical character recognition technique. We implement prototype and evaluate computational cost and filtering performance of the proposed scheme and we attain low false negative ratio with reasonable additional computational cost.