"Nick" <127.0.0.1> wrote
>
>
><(E-Mail Removed)> wrote in message
>news:(E-Mail Removed).. .
>>
>>
>> The above is OK so long as you are not running a business where you
>> want to be 100% sure that a previously unknown potential customer's
>> email will reach you.
>
>If they are sending mail from known spammer addresses, dynamic IPs, etc, I
>wouldn't want their mail. Spamhaus.org Sorbs.net etc make short work of
>killing that crap.
With respect

I am glad you are not in charge of my business
communications, for I would not want you to tell a perfectly decent
potential customer of mine (whose email system happens to be not
favoured by Spamhaus etc) that I don't want his £10,000 order.
The reality of email is that, for any international business which has
a properly managed customer interface and thus *wants* to hear from
*all* genuine people, traditional spam filtering is not a solution.
On a tangent, one can achieve a relative freedom from spam very
easily, by having fairly non-obvious sendmail usernames e.g.
(E-Mail Removed). The spam issue arises if you want to
publish something general like
(E-Mail Removed). I don't actually do
that but I do have sales2@ on the company website, as a graphic (thus
vulnerable only to harvesters who use OCR, or to contacts with
trojan-infected PCs which is something that happens to all normal
people eventually).
The domain I mentioned in my original post has just one username which
causes the spam problem, at a rate of about 1000 per day. The username
is a fairly standard first name, and I can't really get rid of it, or
add "2" to it. If I didn't antispam it somehow, I would get the 1k
spams sent to it and then would be relying on Mailwasher etc which I
know is crap because I did that for quite a while.
I do actually publish an email address on a website on that domain,
for the initial contact, presented as a graphic, and that one is not
antispammed. The username is a name ending in a 3 digit number and I
find it lasts for about a year at a time, which is fine.
Blacklists like spamcop are not good either. We have found that while
hooking sendmail into spamcop reduced the TMDA challenges by about
75%, which *is* good, it also causes a lot of genuine senders to be
blocked. In particular much of say South America in habitually
blacklisted by spamcop, which is really crap but it perhaps reflects
the arrogance of so many Western and esp. American firms. And one
cannot implement a whitelist because the spamcop filtering works at
the sendmail connection level and there is no way to whitelist the
sender before that point.
>> All filters that are effective in blocking today's fairly clever spam
>> will also lose a % of emails.
>
>Anything that looks like spam, that makes it through the blocklists, is
>checked by Spamassassin & the Bayesian filter, & if they fail, they are
>dumped in a spam folder, so can be recovered if needbe.
I'd like to know how an email offering v*agra, with 4 lines of
genuinely sounding text content, and with the product offering totally
within an attached GIF, is going to be identified as spam. You would
have to OCR it, but even then the spammers use corrupted words.
We found, at one point, that dumping all emails that have a GIF
attachment (a trivial exercise, by scanning the email for the MIME
encoding keyword) would be extremely successful. The slight problem
was that a lot of people email on the "company letterhead" which ism
you've guessed, a GIF!
>> I used to run Mailwasher with about 50 filters which one could load
>> into it, and it dumped quite a few legit emails.
>
>The ONLY dumped legit emails resently, were from misconfigered server, i.e.
>one firm had a mail host called wxyz.corp instead of wxyz.com
There are loads of such firms, probably. Very few IT managers know
what they are doing, down to this level. But I still want their
business.
>> Spam filters are OK for personal users who don't care if the
>> invitation for a beer is dumped, and they are OK for anonymous large
>> corporations who don't reply to most emails anyway.
>
>See above. Emails get through.
>
>> But for a small business it is a poor solution.
>>
>> Thank you all for the tips - I am passing them to the chap who looks
>> after the email server.
>>
>> One thing we haven't got is something that recognises an email attack
>> i.e. more than say 10 emails from the same IP within the past hour,
>> and blocks that IP for say 1 week. Apparently, there is no easy way to
>> implement that. But it may be worth doing. The thing is it can't be
>> done in the server firewall because the server also does www. It would
>> need to be a sendmail plugin.
>
>God, what crap are you using. My server software does, mail, ftp, www,
>webmail. It blocks IPs on mail if they try directory attacks (more that a
>certain amount of tries on an IP), blocks ftp if 10 incorrect password are
>tried.
Can you tell me what tool you use and I will pass it on. I'd like to
see this done myself.
>On challenge/confirm, who/where do you send the challenge to ??
To the From: address. Yes, not very satisfactory. But about 99% of
those are fake anyway, and don't get delivered anywhere. So the number
that get delivered to *real* (but forged) addresses is perhaps dozens
per day, which should not be a problem.