Policy Patrol is designed to work with Microsoft's Exchange or Lotus's Domino servers. Management is performed through an MMC snap-in and looks similar to the console for MIMEsweeper.

Rules are built in a similar way, too. Components of rules, including filters, users and actions, are used to build new rules.

For example, once the components had been created, we managed to create a new rule that blocked all outgoing emails that contained words relating to our "Top secret" project.

From this point of the view, the software is incredibly simple to use. But while creating the occasional rule like this is fine for company-specific rules, blocking spam needs to be a little smarter. This is where the Bayesian filtering comes in.

This rule uses real emails – legal ones in an allowed list, spam in a blocked list – to scan email and give it a probability rating that it is spam. All you have to do is choose the threshold value.

The software comes with more than 3,000 spam emails in the blocked list, but it does need you to fill in the allowed list. Fortunately, it is easy to create a rule to add all sent mail, which shouldn't be spam, to the allowed list. You can also add detected spam back into the filter, so that it learns. Using a similar premise, you can configure a new email address that users can send spam too, so you have human eyes improving your spam filtering.

Policy Patrol can also be set for realtime blocking using the lists provided by Spamhaus.org.

But it's not just a product concerned with blocking spam. It can also be used to reinforce company policy and rescue bandwidth. For example, you can compress all attachments over a certain size, and add signatures to all outgoing emails.

Policy Patrol is clearly one of the most powerful tools on test here. The Bayesian filters get better with use and also take into account legitimate mail, while the support for virus scanners and email policy enforcement mean it is much more than a simple spam filter.