Filtering Technology and Blocking Pornographic Web Pages with Images.

Concerns have been raised in [one area] about the effectiveness of web filtering technologies. Specifically, questions have been asked about the inability of web filters to look at web based images. Some have suggested that the lack or inability of web filtering tools to scan images or that employing pixel recognition technology invalidates web filtering. We propose that pixel recognition is not needed in creating an effective web filter. Furthermore, state-of-the-art pixel recognition technology would harm the performance of web filters.

Currently web filters are divided into two categories: list based (passive) and content scanning (active) filters. A list based filter relies on a frequently updated list of inappropriate websites to determine if a web page/domain is appropriate or not. If a webpage address is found on a list/database the web page/domain is inaccessible. Current list based filters have numerous categories and millions of web site URL’s are listed as inappropriate. Creation of the database and updating the list is accomplished by human and automatic review of millions of web pages.

The second more recent and accurate technology is active filtering. As opposed to having a list of web pages to compare a requested web page, these filters actively scan the web page in several methods to classify the web page and determine if the web page is appropriate or not. Filterlogix uses Artificial Intelligence (AI) technology to analyze web pages. We’ve modeled our filter to simulate the way a person comprehends a webpage and determines if the webpage is appropriate or not. In effect, our filter mimics human cognition. We have great success utilizing this technology to prevent pornographic web site access.

The online pornography industry is one of the few web based content providers that are profitable. Each year the industry earns billions of dollars selling access to pornographic web sites. The industry offers many genres of content to satisfy any conceivable taste, from the mild to the extravagant. To remain profitable the industry uses very aggressive marketing tactics to attract new users or “victims”. Because of the highly addictive nature of the subject matter, many surfers are addicted to pornography. This addictive nature further assures a healthy bottom line for pornography providers.

The online pornography industry is a business. Companies must earn money from user subscriptions to pay for bandwidth, content and advertising. Therefore these online content companies must advertise their web pages as well as create web pages that are properly identified by search engines as pornographic. Otherwise these web sites would quickly loose customers and become insolvent. Furthermore, most of these web sites are linked to other web sites via content distribution channels. A handful of companies provide the bulk of the pornographic content to most of the websites.

The above stated facts allow filter companies to create methods and processes to identify and block pornographic websites in varying degrees of success. The Achilles’ heal of pornographic web sites is visibility. The web sites must attract new users through visibility and marketing to generate revenue.

Network administrator who need to acquire content filtering solutions frequently ask filter companies about the possibility of pornographic “stealth site” that users access but filters cannot detect. A typical stealth site might contain images and no text. Hence, conventional filtering technology would be ineffective at stopping these pornographic web sites and the only “solution” to this problem would be pixel recognition technology. While this concern is valid, very few stealth sites exist. Only a handful might be considered marginally successful.

Very few stealth websites exist because of several reasons. The primary reason is the cost of running a website. Web sites need bandwidth and content. Both of these components are expensive. Image intensive web sites need bandwidth requirements beyond the reach of most people. Even high speed internet users have inadequate bandwidth capacities. A successful pornographic web site needs multiple T-1 network connections to the Internet. In most cases pornographic websites use DS-3 connections. These access lines cost many thousands of dollars a month in usage fees. The second component of a pornographic website is material or in this case images. Most images are copyrighted and protected by intellectual property owners. If a foolish webmaster would decide to use copyrighted material without the permission of the owner, legal action on behalf of the owner would be swift.

Pixel recognition technology is seen by many as a panacea to the plight of online pornography but, we must separate fact from fiction. The current state-of-the-art pixel recognition technology is inadequate in dealing with determining the actual subject matter of images. Pixel recognition software is only effective if constraints exist to point the software in the right direction. The Internet has no content constraints and web surfers can look at the whole spectrum of images. Only one outcome can result from this situation, a very high degree of classification inaccuracy. In the past, web filter companies tried utilizing pixel recognition technology however, in independent tests the accuracy rates were abysmal – 5%.

The computation complexity of useful pixel recognition software hinders the technology’s application in any meaningful web filtering capacity. Current generation computer CPU’s do not have the capacity to effectively scan an image and determine with any accuracy the image’s subject. CPU’s would have to increase 10-fold in speed to have the capacity to classify web pages even in the simplest of ways in a web filtering environment.

The Filterlogix iCM (Intelligent Content Manager) is the premiere active filtering product. Our false negative rate (under blocking) is less than 1%. Our false positive rate (over blocking rate) is less than 0.3%. These accuracy rates are proof that our active filtering technology is effective at blocking pornography while providing an unobstructed web surfing experience. If a web site would slip past our filter, our powerful logging/reporting tool could show the filter administrator any unusual web surfing pattern quickly and effectively. The identified website can than be placed on a manual web filtering override list, mitigating potential issues.