Bayesian filter

Filtering of email using probabilities of the occurrence of individual words in ham and spam

A Bayesian filter is a spam filter that uses of the probabilities of the occurrence of individual words in ham and spam emails, and then computes a probability that the email containing those words is spam.

The likeliness of specific words occurring in ham and spam differs a lot between users: for instance, medical terminology and stock exchange-related words are commonly seen in spam emails - however, employees in the pharmaceutical or financial industries respectively are likely to see these words in legitimate email. Bayesian filters therefore work best if they are 'taught' by the user as to what is considered spam and ham.

Spammers are known to use various technologies in their attempts to surpass spam filters, such as adding random bits of text or hiding the content of the email in an image. An overview of such technologies can be found in the The Spammers' Compendium on this website.

Related web links

Quick Links

Poll
The Japanese government is reported to have commissioned a 'defensive virus'. Is 'defensive' malware ever a good idea?
Yes
No
I don't know
Leave a comment
View 11 comments

99 Subscription Promo

Virus Bulletin
In this month's magazine:
  • Living the meme
  • If Svar is the answer...
  • Static analysis of mobile malware
  • And the devil is six: the security consequences of the switch to IPv6
  • Behind enemy lines: reporting from the CCC 28C3 Congress
Virus Bulletin 02 2012
Subscribe now!

Virus Bulletin currently has 224,240 registered users.