How filters work to keep surfing clean

More than two dozen filtering products crowd the market these days in response

to increasing demand from schools, libraries and private users. Though the

products all work to block objectionable content, most differentiate themselves

with various options and levels of protection.

First-generation filters either stopped Internet pages from loading

with manually created databases of sites to avoid, or rigid lists of banned

words. One approach missed too many sites, while the other let too many

through.

To keep up with Internet sites that proliferate at a rate of several

thousand each week, filter makers now say they use software with advanced

artificial intelligence to mine the Web for text streams. Those that fit

certain criteria are automatically added to a list of sites to filter.

Several industry officials say the algorithms built into their software

can analyze words in context so that, for instance, the word "breast" does

not block access to breast cancer information. And they claim teams of editors

review the flagged sites to make sure innocent material is not caught in

the electronic net.

The newer software allows content to be screened page by page to preserve

access to a site where some pages are desirable and some aren't. Customers

can decide which content categories — sex, violence, drugs, hate or even

sports — to filter.

Some services can block chat room discussions, e-mail and newsgroups,

or let subscribers customize their approved or restricted lists and adjust

filtering levels for users of different ages. WebClean, from Lernout &

Hauspie, gives customers the choice to substitute characters for offensive

text.

Filtering software housed on a network server generally is more sophisticated

than the versions people load onto home computers. At some institutions,

Internet traffic is routed through a network hub or a central server managed

by the filter maker or an Internet service provider. Others use an on-site

or proxy server at each location.

When someone tries to access a site, a short query asks the central

computer if the site is appropriate. The server compares the address of

the requested page to a list of pre-screened addresses. If the site is

appropriate, the server retrieves the address. If a site is blocked, the

user will generally get a blank screen with an error message indicating

the site has been blocked.

Companies often download updated lists on a daily or weekly basis.

Many companies offer filtering as part of larger security and Internet

management software packages. To keep library patrons from squatting on

a computer or to help students stay on task, officials can plug in software

that turns off the computer after a set amount of time.

Monitoring and tracking software, such as Cyber Sentinel by Security

Software Systems Inc., allows network administrators to capture screen views

and log addresses users visited on the Internet.

The makers of Cyber Sentinel tout features such as real-time monitoring

of e-mail and instant messaging and the ability to recognize, block and

capture messages from predators and pedophiles. The software can be combined

with another product that displays and forces a user to accept an organization's

acceptable use policy before using a specific application.

N2b Inc.'s Bess Filtering Service, named after the inventor's dog,

and other programs include built-in caches of popular, approved sites. By

storing Internet protocol addresses, a browser doesn't have to wait for

sites to download, reducing network congestion and increasing access speeds

without expensive bandwidth upgrades.