Question
I have used the Bayesian filter tools to mark a message as good but it still shows as junk or spam mail. I would expect any message I mark as being good mail to always appear as such. Why is this not the case?
Answer
Bayesian based filters calculate the probability of a message being good or bad based on the contents of that message. Unlike simple content-based filters, Bayesian filtering learns from both good and bad messages, resulting in a very efficient, self learning, anti-spam system that will return very few false positives.
By marking a message as good you are adding the characteristics of that message into the Bayesian dictionary. This does not guarantee that the message will be regarded as good the next time it is viewed/received.
Each time a message is viewed/received the Bayesian filter takes the message, runs comparisons against its dictionary of good and bad messages and assesses the probability of the message being spam or not spam based on this information.
As the dictionary can contain information on thousands of messages of each type, the addition of one more message will not hugely affect the probability of that message being one thing or the other, unless there is already a reasonably similar message profile in the dictionary.
See Also:
- Whats new in Build 3180?
- Which new viruses are caught with the GMS Anti-Virus definition files released on 06 Jul 2006?
Keywords:Bayesian filters good junk spam