Release Notes: This release adds dspam_signature, a tool for decoding DSPAM Signatures for debugging or personal interest.
Release Notes: Base64 decoding of text and HTML segments (to catch encoded spams), a button for the exceptionally rare false positive from a quarantine box, and filtering of short HTML comments (used to frequently break up words in spams) were all added. CGI code was cleaned up, fixing a few bad redirects to index.cgi. DSPAM statistics were added to CGI. The DSPAM signature, pasted onto every email with reversal information, was changed to a base64-encoded, delimited signature similar to a PGP or cert signature. Old reversal information is automatically stripped from replied-to email (it does not strip quoted).
Release Notes: A small fix has been made to the Linux makefile (-ldb -> -ldb1). Lines beginning with 'From' are prefixed when writing to the quarantine box. A bug that resulted in caught spam messages never showing up in the quarantine box has been fixed (they were lacking the initial "From").
Release Notes: If referal information is found in a spam message, the original 15 tokens will be used for both reversal and spam calculations (having found that the original 15 tokens were generally more useful than a second set of tokens generated). This release ignores the recipient's header for messages forwarded as spam that don't have reversal information and pays more attention to the original forwarded header. A bugfix improves the embedding of reversal information into multipart emails.
Release Notes: Many calculation changes have been made. Reversal information is now embedded into every email so that the previous token statistics can be reversed if the email is spam. Tokenizing has been enhanced to take uppercase words into account as well as header information.