256563 - Implement junk-filtering based on the Chung-Kwei (Teiresias) algorithm

Status: NEW

(MailNews Core :: Filters, enhancement)

Description

[Simon Paquet [:sipaq]]

The Thomas J. Watson Research Center <http://www.watson.ibm.com/> has done some
pretty interesting research on using the Teiresias algorithm (which was
originally used to identify similar DNS sequences) for filtering SPAM.

The research paper can be found at the given URL. The first preliminary results
are pretty impressive:

- 96,56 % success rate on a SPAM body of 65000 messages.
- only 1 false positive out of 6000 messages.
- Scan rate of 214 messages per second on a 2,2 GHz P4.

Combined with our current Bayes-based SPAM filtering. This could get us even
better results than now and increase our reputation as the best client for
SPAM-filtering.

Comment 1

[Simon Paquet [:sipaq]]

Moving to the default assignees, since Scott is clearly not working on this. Perhaps that will help to find someone to eventually take this bug.

Comment 2

[Matt]

I have updated the link to the scholarly paper. Short as it is.