some bogopass (base64) results
David Relson
relson at osagesoftware.com
Wed Nov 27 02:38:11 CET 2002
Suzanne,
I got bogopass working ... finally. Turns out using Eudora to read mail
and process attachments (for example bogopass) and using samba to move them
to the linux box isn't quite good enough. bash includes the '\r' on the #!
line when it tries to run perl. As I don't have /usr/local/bin/perl (a
symlink to /usr/bin/perl), but don't have a /usr/local/bin/perl\r, bash was
complaining. Having figured that out, bogopass works fine and I have a
couple of test results to report.
First, there are my two test files. Both are spam with normal headers and
the whole body encoded as a single block of base64.
-rw-r----- 1 relson relson 14527 Nov 19 19:43 base64.1.txt
-rw-r----- 1 relson relson 1998 Nov 25 16:05 base64.2.txt
I ran each of them through bogofilter (generating $file.bf) and through
bogopass (generating $file.bp). Here are the spamicity scores:
w/o decoding with decoding
base64.1.txt 0.603784 0.572895
base64.2.txt 0.715739 0.549387
It appears that just evaluating the headers (as happens without base64
decoding) gives a higher spamicity score. As a guess, the decoded message
contains lots of ordinary words which lower the score.
David
More information about the Bogofilter
mailing list