Bayesian filtering - Email Policy Controls

Email policy controls give the administrator even greater scope to create more extensive rules to deal with email, customized to your organisation's needs.

The Email Policy system can also work in conjunction with other UTM services which deal with email - in particular Sophos powered anti-virus scanning and anti-spam. When either of these two services discovers an incoming (or outgoing) email which is deemed to be infected with a virus or is spam, they can quarantine the email, which is where the Email Policy system can take over.

Additional rules and controls used by the Administrator can then be initiated. The Email Policy can also be actively employed as a second tier of checks. For example after (or even before) an incoming mail has been checked for the usual spam characteristics you could decide that any email coming from a known source (say from one of your customers) would always be treated as a ham (good) message, regardless of the content. You are creating what is known as a whitelist. The exact opposite could also be set up, say any emails coming to your staff from your competitors could always be deleted (blacklisted) regardless of the content being deemed spam or whether it caries a virus or not !

Email Policies can also be used to control size of emails, create footers, create out of office responders, check content for inappropriate content or block particular users from gaining external email access.

Bayesian filtering is a mathematical approach that, unlike many other anti-spam technologies, adapts over time and takes the changing strategies of spammers into account. It therefore offers moving goal posts which obviously make it far harder for the spammers.

Central to Bayesian filtering is the principle that the likelihood of events happening in the future can be inferred by analyzing past events. Spam emails are therefore likely to be made up of similar elements, while valid emails (sometimes referred to as 'ham') will have their own determining characteristics. Bayes classifiers learn as they go, updating both the rules and the scores. When a new evasion trick comes along, the message may still have enough other bad features that the filter will recognize as spam. If so, the system will automatically learn the new spam characteristics.

Spam assessment read more>>

Spam Cop read more>>