SpamBully
http://www.spambully.com/ is very good. Works nicely with
Outlook. Filters and has some other decent features.
The OP asked for a *FREE* anti-spam solution (read the Subject header in the
OP's post). SpamBully is NOT free. It costs $30; see
http://www.spambully.com/register.php.
Like Mailwasher and its bounce option, SpamBully is yet another that
provides this stupid feature. Spammers don't use their own e-mail address
so the bounces will go nowhere (the domain doesn't exist), or they will
waste resources at a domain (the spammer used a valid domain but not a valid
username), or they hit innocents (the spammer used a valid e-mail address
but which was someone ELSE's e-mail address).
Challenge-Response, especially when implemented at the client, is another
stupid anti-spam solution. It is an irresponsible solution because it hits
innocents with your "challenge spam". Again, spammers aren't using their
own e-mail address so if it is a valid e-mail address then it hits an
innocent, and since the innocent never sent you an e-mail then your
challenge is unsolicited and hence "challenge spam". The easiest way to
thwart these idiots is to respond to the challenge so the bozo that uses it
will get the suspect mail moved into the Inbox and they end up seeing the
spam that they tried to involuntarily enlist you as their spam filterer.
It's not my job to filter out spam from someone else's mailbox. See:
http://kmself.home.netcom.com/Rants/challenge-response.html
http://richi.co.uk/blog/2005/05/why-challengeresponse-is-bad.html
http://www.google.com/search?q=+"challenge-response"+%2Bbad
SpamBully seems to rely only on Bayesian filtering to detect spam. For spam
contained within a GIF MIME part, Bayesian is worthless. Bayesian will
generate false positives because, after all, it is guessing based on word or
phrase weighting, and it ignorant until you have passed enough mail traffic
for it to build a decent database (some allow you to pre-train the Bayesian
filter but SpamBully isn't one of them). I saw no means for the user to
reclassify a message so the database weighting gets changed. A false
positive should be removed as spam (and added as ham) and visa versa. If
you cannot reclassify, your database has invalid records. I also saw no
configuration for reducing "noise" in the database; i.e., expiring
weightings for words that have not reappeared in any mails for a
configurable period of time (if they don't appear anymore, eventually they
are noise and should be treated anew when they do appear). Some spams
deliberately attempt to poison Bayesian databases by upping the noise floor
in their database.
There is mention of blacklists but nothing shown how to configure which ones
to use in their screenshots. I saw an "RBL" tab in one screenshot which
might be for "real-time blacklists". I saw a screenshot at
http://www.spambully.com/sb3help/index.php?page=index_v2&id=18&c=5, so
apparently the user has to know the URL to use to access the various
blocklists. You can't even pre-define a list of blocklists and then use a
checkbox for which ones you want to experiment with at the time. Each one
has a different "flavor" of how it identifies spam sources. You don't want
to use SPEWS because they rate spamminess for domain (spam friendliness,
spam laziness, lack of action) rather than identify the spam sources. SORBS
subscribes to the same ideology as SPEWS but is a bit less agressive but
they are also very slow to update their list. SpamHaus SBL+RBL is good
along with NJABL. I also use SpamCop but they are bit more agressive yet
their list is responsive. Without a choice of WHICH of the DNSBLs (DNS
blocklists) you use, you have no means of tailoring how aggressive is your
spam identification.
There is no detection of mails coming from dynamically assigned IP addressed
hosts (i.e., for users on dial-up or broadband) and that is from where the
mailer trojans spew their turds. Bayesian should be the last method of
catching spam, not the first.
Regardless of [dis]agreement over the features of SpamBully, it does NOT
match the criteria set forth by the original poster, and that was the
product was to be free. SpamBully is not free.
