Generation of pornographic blacklist and its incremental update using an inverse chi-square based method

Mining browsing behaviors for objectionable content filtering

“…Among several self‐rating mechanisms that have been built on PICS, the labeling system promoted by the Internet Content Rating Association (ICRA) is a popular one. Nevertheless, experimental analysis among 10,000 bilingual pornographic web sites showed that only 5.4% used ICRA labels (Lee & Luh, ). Thus, the self‐rating labeling mechanism still has a long way to go because objectionable content providers rarely fulfill this requirement to avoid being filtered easily.…”

“…URL‐based features have been extracted for classifying pornographic web pages (Zhang et al., ). The title of a web page has been empirically shown to be a cost‐effective feature for early decision making on classifying pornographic web pages (Lee et al., ,). An inverse chi‐square‐based method has been employed to mine content features for generating pornographic blacklists (Lee & Luh, ).…”

“…Controlling access to the Internet through filtering has been recognized as a mandatory response to address the issue of content inappropriate for children. Traditional filtering techniques regard objectionable access identification as a categorization problem through content analysis (Caulkins, Ding, Duncan, Krishnan, & Nyberg, ; Chau & Chen, ; Chen, Wu, Zhu, & Hu, ; Deselaers, Pimenidis, & Hey, ; Eickhoff et al., , ; Hammami, Chahir, & Chen, ; Ho & Watters, ; Jansohn, Ulges, & Breuel, ; Lee & Luh, ; Lee, Luh, & Yang, ; Lee, Hui, & Fong, , , ; Lin, Jan, Lin, & Lai, ; Polpinij, Sibunruang, Paungpronpitag, Chamchong, & Chotthanom, ; Wu, Zuo, Hu, Zhu, & Li, ; Zhang, Qin, & Yan, ; Zhu, Wu, Cheng, & Wu, ; Zuo, Hu, & Wu, ), that is, crawling the content of URLs and analyzing the texts, images, video clips, and films by machine learning to distinguish normal web pages from objectionable ones. Unlike the case in most previous studies, which have focused on the content of web pages for objectionable access filtering, we look for insights into users' web surfing behaviors in click‐through data.…”

Mining search intents for collaborative cyberporn filtering

“…They attempt to analyze texts, images, video clips, and films on web pages to understand the context in which the discriminative features appear and make a classification decision. Several numerical algorithms, including neural networks (Lee et al, 2002, 2003, 2005), statistical text analysis (Caulkins et al, 2006; Lee & Luh, 2008; Lee, Luh, & Yang, 2008), pattern‐recognition techniques (Deselaers, Pimenidis, & Hey, 2008; Lee, Kuo, Chung, & Chen, 2007), and hybrid feature analysis (Chen, Wu, Zhu, & Hu, 2006; Hammami, Chahir, & Chen, 2006; Polpinij, Sibunruang, Paungpronpitag, Chamchong, & Chotthanom, 2008) have been used to train the classifier for pornography filtering. Features from skin pixels were usually adopted for pornographic‐image recognition (Wu, Zuo, Hu, Zhu, & Li, 2008; Zhu, Wu, Cheng, & Wu, 2004; Zuo, Hu, & Wu, 2010).…”

“…Most previous studies have regarded cyberporn‐filtering problems as categorization with statically crawled datasets. Lee and Luh (2008) explored pornographic hub sites by web structuring and cost‐effectively mining for updating pornographic blacklists on the changing web. Users usually access web content by interacting with search engines while search engines periodically refresh their indexing, including adding new web pages and removing those which are defunct.…”