Spam filtering has become a growing market in recent years as spammers have become increasingly brazen in their efforts to peddle their wares. Customers have found Microsoft Office Outlook 2003's junk-mail filtering technology, known as SmartScreen and developed at Microsoft Research, to be quite effective despite its limitations (e.g., the filter works only with Exchange Server 2003 cached-mode profiles). However, the Outlook filter is still a client-side tool that works on email only after that mail has been delivered to an Exchange mailbox. At COMDEX 2003, Microsoft Chairman and Chief Software Architect Bill Gates announced that Microsoft is developing a server-side spam filter that integrates with Exchange 2003: the Exchange Intelligent Message Filter.
Microsoft Research specializes in long-term research projects, many of which (e.g., Microsoft ClearType, speech recognition) are incorporated into Microsoft products. SmartScreen filtering technology made its debut in Microsoft's MSN client and was already a proven technology when the Outlook team incorporated it. The SmartScreen filter works by scanning inbound messages and applying heuristic rules to decide which messages are spam and which aren't. SmartScreen implements these rules as a Bayesian filter, which is widely used in UNIX spam filtering and which works well after it's been trained with a sufficient number of messages. Microsoft has been able to train the SmartScreen filter with email that MSN Hotmail users who contributed to the training effort have forwarded.
Exchange Intelligent Message Filter, which Microsoft plans to release early next year, is designed to hit the filtering sweet spot by catching junk mail at the network perimeter. Exchange Intelligent Message Filter is a server-side component that will run on Exchange 2003 and will integrate with Outlook 2003's and Outlook Web Access (OWA) 2003's trusted and junk senders lists to control what senders and domains users do or don't trust implicitly. Exchange Intelligent Message Filter will provide two levels of filtering--a gateway threshold and a mailbox-store threshold--according to a message's spam confidence level (SCL) property. (Third-party filters already use the SCL.) Exchange Intelligent Message Filter will set this property, which rates the likelihood that the message is spam, for incoming email messages. Administrators can set the threshold levels and define certain actions to take place when a message's SCL crosses one or both thresholds.
The gateway threshold controls which messages the filter will stop at the gateway. For example, you can configure Exchange Intelligent Message Filter to delete messages that exceed a gateway threshold of 75 or to quarantine messages that exceed a threshold of 90. Messages that pass the gateway threshold go on to the mailbox store, in which the filter applies the mailbox-store threshold. If the message's SCL exceeds this threshold, the filter automatically moves the message to the user's Junk Mail folder after evaluating the user's trusted sender and trusted domain lists. This approach works quite nicely with both Outlook and OWA and, because the filtering happens on the server, messages that arrive while a user is offline are still filtered at the mailbox-store level.
Microsoft hasn't yet announced final availability or pricing for Exchange Intelligent Message Filter, but more details should be forthcoming in the next few weeks. As soon as I can get a close look at this new component, I'll examine it in detail and report my findings to you.
