Slightly OT: What's a good score for anti-spam software?

[ceed]

Hi,

Free anti spam software is often discussed here. The good thing about anti
spam programs is that most of them provide some form of statistical
information. I would like to know what kind of accuracy people here expect
for them to be able to say they recommend an anti spam solution.
Expectations and results will of course differ depending on how much spam
your email address receives, and what filters mail passes through before
it hits your solution. Still, it would be interesting to know what people
find acceptable regardless of these other factors. Is it 95%, 99% or even
above 99.5%?

I myself receive about 60% spam and 40% legit email in average and would
not recommend any solution that didn't get rid of at least 99% of that
without false positives after 30 days of learning and tweaking.

 

[badgolferman]

Hi,

Free anti spam software is often discussed here. The good thing about
anti spam programs is that most of them provide some form of
statistical information. I would like to know what kind of accuracy
people here expect for them to be able to say they recommend an anti
spam solution. Expectations and results will of course differ
depending on how much spam your email address receives, and what
filters mail passes through before it hits your solution. Still, it
would be interesting to know what people find acceptable regardless
of these other factors. Is it 95%, 99% or even above 99.5%?

I myself receive about 60% spam and 40% legit email in average and
would not recommend any solution that didn't get rid of at least 99%
of that without false positives after 30 days of learning and
tweaking.

K9 is filtering at 99.82% for me. Although that may be considered very
good by others I have had higher accuracy rate up to six months ago.
It has become a bit more aggresive in marking Good mail as Spam and I
have had to change the way it filters. I receive around 130 messages a
day and maybe five of them are legitimate ones, the rest is junk.

 

[ceed]

K9 is filtering at 99.82% for me.

I'm using K9 as well and I'm at 99.89 right now. No false positives the
last 30 days. I was using SpamPal for a while and got the same results,
but had to maintian a lot of plug-ins and tweaked settings. K9 is
wonderfully simple. Not much traffic on the K9 forums though. Maybe that's
a good sign: People have nothing to gripe about?

 

[Mel]

K9 is filtering at 99.82% for me. Although that may be considered very
good by others I have had higher accuracy rate up to six months ago.
It has become a bit more aggresive in marking Good mail as Spam and I
have had to change the way it filters. I receive around 130 messages a
day and maybe five of them are legitimate ones, the rest is junk.

Some Mail Services provide Spam Filtering (mine does), however, some
Spam still gets thru. They say: If I forward the ones that got thru to
them that they will update their Spam Filters to block future Spam from
the same folks.

 

[Richard Steinfeld]

Hi,

Free anti spam software is often discussed here. The good thing about
anti spam programs is that most of them provide some form of
statistical information. I would like to know what kind of accuracy
people here expect for them to be able to say they recommend an anti
spam solution. Expectations and results will of course differ depending
on how much spam your email address receives, and what filters mail
passes through before it hits your solution. Still, it would be
interesting to know what people find acceptable regardless of these
other factors. Is it 95%, 99% or even above 99.5%?

I myself receive about 60% spam and 40% legit email in average and
would not recommend any solution that didn't get rid of at least 99% of
that without false positives after 30 days of learning and tweaking.

I recommend, if possible, switching to an ISP who employs Spam Assassin
on their own dedicated servers. This is a cooperative open-source system
that's the result of pooled talent from a number of _smaller_ internet
providers.

The result for me has been excellent. I receive about 1-2 spam messages
per day and I haven't bothered munging my email address lately. The
spammers usually pick up our email addresses from deeper in the message
headers anyway. So, you may notice my munging right here -- this is
mostly to slow down one or two crazy people on a couple of newsgroups
that I frequent. I do not use any spam solutions on my system because I
don't need to. My ISP intercepts 3-6 additional spams per day
specifically addressed to me, in addition to a barrage of god knows how
many more that are easily trashed at their level -- things like
dictionary attacks, which are very obvious when they arrive in batches.

I've been burned badly with Pacific Bell/SBC, underwhelmed with
Earthlink, heard too many horror stories about AOL. Hell will freeze
before I do business with Yahoo.

Small is beautiful.

Richard

 

[ceed]

I recommend, if possible, switching to an ISP who employs Spam Assassin
on their own dedicated servers. This is a cooperative open-
source system that's the result of pooled talent from a number of
_smaller_ internet providers.

I run my own mail server (Linux) and use Spamassassin 3.x on it. Not all
is caught though. At server level false positives are much more of a
problem than after the mail reaches it's destination, so Spamassassin
can't be set to operate as strick as client side applications can.

The result for me has been excellent. I receive about 1-2 spam messages
per day and I haven't bothered munging my email address
lately. The spammers usually pick up our email addresses from deeper in
the message headers anyway. So, you may notice my
munging right here -- this is mostly to slow down one or two crazy
people on a couple of newsgroups that I frequent. I do not use any
spam solutions on my system because I don't need to. My ISP intercepts
3-6 additional spams per day specifically addressed to me,
in addition to a barrage of god knows how many more that are easily
trashed at their level -- things like dictionary attacks, which are
very obvious when they arrive in batches.

I've been burned badly with Pacific Bell/SBC, underwhelmed with
Earthlink, heard too many horror stories about AOL. Hell will freeze
before I do business with Yahoo.

Small is beautiful.

Sure is. That's why I run my own mail....

 

[Aaron]

<ceed@abcdefghijklmnopqrstuvwxyzabcdefghijklmnopqrstuvwxyzabcdefghijk.com

wrote in newsp.stwormbc21xk10@dellbob:

I'm using K9 as well and I'm at 99.89 right now. No false positives
the last 30 days.

One thing to note is that acccuracy scores like this, don't tell the
whole picture. 2 Spam filters might have the same accuracy rate say
99.5%, but for one it might be misindentifying a lot more ham as spam and
have a relative high False positive rate.

This is espically so if you generate a lot of spam mail in porportion to
ham mail say 50 to 1. In such a situation, you can easily score an
overall accuracy of 99.5% and up, when the filter correctly indenitifies
spam as spam but still have a high false positive rate because it thinks
wrongly that ham mail is spam.

http://www.jgc.org/astlt/ talks about spam hit rate and ham hit rate.

# The Spam Hit Rate is the fraction of spams correctly identified
# The Ham Strike Rate is the fraction of ham messages incorrectly
identified