Skip to main content

The federal government's omnibus crime bill, with its harsh penalties against child pornography on the Internet, is being praised as a good first step for Canada. But a small Internet company in Guelph, Ont., has been doing something about it for the past three years.

NetSweeper Inc. is a World Wide Web filtering service for large organizations that boasts it can block up to 99 per cent of unacceptable Web content. And its success is spreading - although the company has only 20 employees, it says it has a presence around the world, including offices in Panama, Britain, South Africa, the United States and Peru.

In Toronto, one of its clients includes all 129 branches of the Toronto Public Library.

While most attempts to block sites containing pornography, hate material or hacking tutorials rely on software installed on an organization's network server, NetSweeper is more direct: it maintains a databse of Web addresses tailored to the client's wishes and needs.

It's a service, explained co-owner and chief technology officer Lou Endelyi, "an almost plug-and-play solution."

A NetSweeper client leases a hardware filter that sits between the open Internet and the client's network. The filter intercepts outgoing requests and matches them against a database of Web sites the client wishes to block, and sends a message back when users request a banned site.

The system is quick, said Mr. Endelyi, because it is essentially a proxy server with a fast central processor, as much memory as a network demands, and it does not interfere with any other Internet activity, such as serving e-mail (which requires a lot of processing power). And because NetSweeper concentrates only on outgoing requests, it does not waste time examining pages as they come in and trying to block them according to key words.

The heart of the process, he said, is that NetSweeper is constantly looking for sites that fall into a growing list of categories of questionable content. It stores their addresses or IP numbers in a database on a cluster of computers connected via fibre-optic cable to the Internet. The company then sends database updates daily without waiting for the client to request them.

Software filters alone can't do the job properly, said David Young, a marketing consultant working with NetSweeper, because software is geared to key words, and thus blocks some respectable sites.

"Say you ask for a recipe containing a breast of chicken," he said. "Software could stop it from coming in. The same thing would happen if you requested the home page of the Middlesex County School Board."

NetSweeper categorizes each page according to content, not key words, and its technique for doing this is the company's proprietary secret, said. Mr. Endelyi. But he can say what the process is not.

"It's not artificial intelligence, because NetSweeper can't grow on its own," he said. "And it's not fuzzy logic, because NetSweeper is actually looking for certain information. In fact, it's much more efficient than any search engine."

Because clients define different kinds of content they wish to block, NetSweeper breaks down its objectionable material into customized databases. These databases are tailored from a list of categories - at the moment, there are more than 30 - that NetSweeper will further refine in a dialogue with the client. "We have a human quality-control program," said Mr. Endelyi, "so we can always check our effectiveness."

For instance, schools may wish to block pornography, hate speech and hacker sites, while corporate owners may be more concerned with employees spending their time filling out forms on résumé sites, and libraries will want to stop visitors from playing on-line games. Some clients, said Mr. Endelyi, will wish to block out only parts of some Web sites - for instance, all of the IBM Web site would be accessible except that part in which one could purchase goods on-line.

The company has a method of stopping e-mail, but only mail that operates as a purely Web protocol, such as Hotmail and Yahoo. NetSweeper says it plans to extend its service to all e-mail and Usenet newsgroups some time in the future. But the Web remains a high priority, Mr. Endelyi explained, because 95 per cent of unacceptable content comes from the Web, while only 4 per cent arrives by e-mail and 1 per cent by all other methods, including Usenet and Internet Relay Chat groups.

Just keeping up with the Internet's expansion is keeping the company busy, he added.

"There are 25,000 new top-level domains being registered every day," he said, "and about 2,500 disappear daily, so we have to work 24/7 to keep up."

Mr. Endelyi sees nothing but growth ahead.

"Internet users are doubling every 89 days," he said, "and over the next three years another 1.6 billion users will go on-line.

"Net filtering is the next frontier."

Interact with The Globe