"Mark" said in news:%
[email protected]:
The premise of which is that if a static list of unweighted words
doesn't work then use a Bayesian filter that statistically weights words
by their historical appearance in spam and non-spam messages. That's
okay but does require that you actually download the message so the
Bayesian filter can look at it to interrogate the words used within it.
Check if your e-mail account with your ISP has an option to enable an
anti-spam filter. If so, turn it on. Proactive server-side solutions
are better than reactive client-side solutions. Then decide if all you
want for anti-spam prevention is what minimal features Microsoft might
include in their products. Have you yet bothered to investigate what
anti-spam products there are available to you?
Norton's SpamAlert (comes with their firewall product) or their AntiSpam
standalone product might work for you. It costs money and I don't feel
they are as capable as other products from what I've seen of their
SpamAlert product (I use their firewall - for now). McAfee has their
SpamKiller, and it also costs money. SpamAI costs money, too. There
are lots of anti-spam products that want to make a buck off the victims.
K9 is another Bayesian-only solution and is free but I haven't used it
so I cannot vouch for its effectiveness. Most of the anti-spam
solutions rely on looking at only the content of the message (which
means they have to download it). Some other anti-spam solutions also
check *where* the message originated and will detect spam that comes
from known sources of spam. I don't know all of these products but
MailWasher and SpamPal come to mind.
Mailwasher has free and paid versions. The paid version includes
support for Hotmail and multiple accounts. Their freebie only supports
one account (according to Mailwasher's web site) and has no support
other than POP3. Only recently did I see mention in another newsgroup
thread that Mailwasher supposedly has an option buried under its
blacklist options to automatically delete from the mail server any
messages suspected of being spam. I don't know if it also includes a
log of those automated server-side deletes so you can monitor for false
positives. Mailwasher uses the publicly available blacklists which list
known spam sources. I don't know which blacklists to which Mailwasher
will subscribe, but some can be overly aggressive, like SPEWS, and cause
too many false positives. Other than using blacklists, I don't know if
Mailwasher supports other means of detecting spam, like a Bayesian
filter, looking for HTML tricks, checking for a missing plain-text part
in a MIME encoded message that has an HTML part. Spammers using
HTML-formatted messages also very often violate multipart MIME and do
not provide a plain-text MIME part. Some webmail providers also violate
multipart MIME, like Hotmail (so you have to whitelist this domain or
specific senders from there if you block MIME encoded messages that have
no plain-text part). I really doubt that Mailwasher checks the URLs
contained within a message to see if they go to a known spam site. I
use SpamPal to provide me all of those features.
SpamPal is free. There is no paid-for version although you can donate
if you want to further the cause. It isn't nagware, crippleware,
trialware, demoware, bannerware, or any other crapware type. Lots of
other altrustic authors have coded plug-ins to enhance SpamPal.
SpamPal, like Mailwasher, itself detects spam based on *where* it
originates knowing that most of the spam comes from these sites. It can
also block messages based on their country of origin, so if you live in
the USA then you probably don't expect nor want e-mails from China,
Argentina, or Malaysia (and visa versa for folks over there that don't
want nor expect e-mails from the USA). Not every country is listed,
just the major offenders. Besides a whitelist feature to let good
e-mails pass through, it has an auto-whitelist feature to monitor who
sends you e-mails and to whom you reply.
There are lots of plug-ins to extend SpamPal to include checking of the
message content should you desire to increase the chances of detecting
spam. There is the Bayesian plug-in to give you the same ability as
OL2003 (without having to pay for the upgrade) and found in other
content-only anti-spam filtering products, like K9. This weights a word
list based on their past usage in spam or non-spam e-mails, plus you can
edit an exclusion list. The URL plug-in will check any URL links within
the message to see if any go to spam sites. The HTML-Modify plug-in
will detect spam based on tricks that spammers use, one of which is to
provide only an HTML part of the message but not a plain-text version to
make it harder for recipients to see the real message instead of having
to render it from its HTML coding. You can also have the plug-in strip
out linked images which are employed by spammers as web bugs to see when
you have opened their e-mail. Embedded images are retained (unless you
want them filtered out, too). A RegEx plug-in lets you add more control
by defining regular expressions that go beyond what you can usually
define by rules in an e-mail client to catch spam, but you'll need some
education in how to write regular expressions. The only real deficiency
to SpamPal is that it is passive and simply marks what it detects as
spam. You must define a rule in your e-mail client to decide what to do
with the spam-marked message.
SpamPal is not proactive in deleting the spam off your mail server so
you never have to bother wasting the time and bandwidth to download the
message where you then detect and delete it locally. You could define
your e-mail client's rule to delete the message but that is a local
delete. You could define your e-mail client to only download the
headers and use SpamPal's primary defense of detecting spam based on
where it came from and have the rule delete the message (which would
then perform a delete on the mail server). I've never considered
Outlook stable enough to leave running when configured for only
POP3/SMTP accounts (it's okay in a corporate environment when using
Exchange). So I use Magic Mail Monitor (mmm3.sourceforge.net) which is
just an e-mail monitor that I can leave running all the time but it also
has rules you can define. I tried PopTray for awhile but its header
rule often does not fire. In Magic, I define a rule to check if SpamPal
marked a message as spam; if so, Magic's rule then deletes that spam off
my mail server so I'm never afflicted with it. Since SpamPal can work
on just the headers, it is not required to download the whole message
for Magic to see SpamPal detected the message as spam. Downloading just
the headers is much faster and less burdensome on your network
connection then having to download the complete body of the message.
Magic is free, too.
So my setup is:
e-mail client <-- SpamPal <-- POP3 server
where the e-mail client is Outlook, Outlook Express, or my Magic monitor
(which also has rules to do server-side deletes). Magic includes a log
of all deletes performed by its rules along with showing the standard
headers for the deleted messages so you can check for false positives
(i.e., messages that weren't spam based on your opinion separately of
how you configured SpamPal).
Regardless of what anti-spam solution you choose, you'll probably want
to define some overriding rules in your e-mail client. For example, one
of the topmost rules should be to keep e-mails from known senders (i.e.,
those in your address books). It is highly unlikely that you will keep
a spammer listed in your address book. If the anti-spam product doesn't
include a whitelist function, you'll need to define some whitelisting
rules positioned at the top of the rules list to keep those e-mails,
like those sent from newsletters you have subscribed.
Lots of anti-spam products work by providing a Bayesian filter. So does
SpamPal (with a plug-in). Some work by checking from where the message
originated. That includes SpamPal. Some will check the URLs in the
message to see if they go to a spam web site. The only one I've heard
that does this is the URL plug-in for SpamPal (but it still needs
additional work to overcome some of the trickier obfuscation methods
employed by spammers to hide the target of a URL). Some let you define
rules. SpamPal has its RegEx plug-in, plus you can define rules in
Magic which can even perform server-side deletes while only checking on
the headers. Some let you block based on domains. SpamPal has its
blacklist features and tis easy 'nuff to specify something like
"\.<ccTLD>$" in the RegEx plug-in to block on specific country codes
(where <ccTLD> is the country code top-level domain, as in
domain.com.ru), but it really helps to use the country blacklist option
in SpamPal to catch those that don't use a ccTLD.
I was thinking at one time of using Mailwasher but their freebie version
only supports one account. I have 6. At the time, no one could tell me
that it could perform automatic server-side deletes (other than the
expect mode of operation where the user still has to choose which ones
to delete). I really don't want any product that supports itself by
aggravating me with banners. I think the Mailwasher banner is probably
now disabled or unused but it still occupies a chunk of screen space. I
like the altruistic attitude of Spampal's author and those of its
plug-ins in trying to help the community eliminate a problem rather than
trying to profit from it. Mailwasher's (the freebie version) goal is to
get you to buy their paid-for version. SpamPal's defect was that it is
passive, but adding Magic makes it active. With SpamPal and Magic, I'm
pretty well armed against spam. I'm always looking for something
*better* (not just equal or somewhat less capable). Mailwasher isn't
it, neither are K9 or SpamAI, and I have yet to hear a decent threshold
of good comments regarding Norton and McAfee products (but it is also
hard to beat a price of FREE).
About the only negative to Spampal is that it runs as a local proxy.
Not all users are familiar with using a proxy or they are too timid to
try one, but setting up SpamPal is pretty easy. The hardest part is to
reconfigure the e-mail account so the POP3 server points to SpamPal and
to edit the username to include both the username along with the POP3
server name from which you get your e-mails; i.e., "joeschmuck" becomes
"
[email protected]". However, I don't see this learning
curve any steeper than the newbie that has to figure out how to create
and configure an account in their e-mail client.
