Maybe it's because I'm filtering my SPAM for later manual "safety"
screening as opposed to deleting it completely, but it seems like some of
the algorithmically generated scores just pussy-foot around too much.
Here's are scores for a recent slip-through, a common pyramid scheme message:
X-Spa
> "RW" == Rob Winters <[EMAIL PROTECTED]> writes:
RW> RAZOR_CHECK version=2.20 = 3.0 (a manual score, soon to be 5.0)
Before you give razor the ability to block your mail all by itself,
consider the false positives from mailing lists. Apparently there are
some fools out there that are inten
On Wed:11:43, Rob Winters wrote:
[...]
> SA does not give any credit to the cumulative effect
[...]
It seems to me that properly weighted scores would
avoid this problem. I'd like to think that a good
optimization algorithm, such as a genetic algorithm,
could do the job.
Thanks,
Kingsley
__
On Wed, 29 May 2002 the voices made Kingsley G. Morse Jr. write:
> On Wed:11:43, Rob Winters wrote:
> [...]
> > SA does not give any credit to the cumulative effect
> [...]
>
> It seems to me that properly weighted scores would
> avoid this problem. I'd like to think that a good
> optimization al
On Wed, May 29, 2002 at 08:45:41PM +0200, Tony L. Svanstrom wrote:
| On Wed, 29 May 2002 the voices made Kingsley G. Morse Jr. write:
| > On Wed:11:43, Rob Winters wrote:
| > [...]
| > > SA does not give any credit to the cumulative effect
| > [...]
| >
| > It seems to me that properly weighted sc
Thats why in spam assassin you can set the scores yourself... fit them for
your needs..
Brian
- Original Message -
From: "Rob Winters" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Wednesday, May 29, 2002 8:43 AM
Subject: [SAtalk] large numbers of tiny score
Good point. Combinations of some rules may be more
indicative of spam than others.
It would be great if the GA could infer the boolean
logic, as well as the scores.
Thanks,
Kingsley
On Wed:20:45, Tony L. Svanstrom wrote:
> On Wed, 29 May 2002 the voices made Kingsley G. Morse Jr. write:
>
> >
At 03:23 PM 5/29/2002, Brian May wrote:
>Thats why in spam assassin you can set the scores yourself... fit them for
>your needs..
Well, adjusting the scores won't necessarily make the tool better. I'm sure
that the computationally-derived scores are excellent. In fact, I submit
that you've pro
Wouldn't it be possible to add this as just another test in the GA?
A rule that looks at all the previous rules that matches. Just make sure
the GA doesn't do anything with it until the other rules are calculated.
Some percentage of the score could be the multiplication factor that is
used.
On
Rob Winters wrote:
> How about a "bonus" for cumulative effect? Why not do a second-level
> analysis after scoring; something like:
>
> 3 positive score matches - add 1.0
> 4 positive score matches - add 2.0
> 5 positive score matches - add 4.0
> 6 positive score matches - add 8.0
This reminds me
On Wednesday, May 29, 2002, at 08:43 AM, Rob Winters wrote:
> How about a "bonus" for cumulative effect? Why not do a second-level
> analysis after scoring; something like:
>
> 3 positive score matches - add 1.0
> 4 positive score matches - add 2.0
> 5 positive score matches - add 4.0
> 6 positi
> I came up with the name "Five-Card Charlie", which is a reference to the
> game of Blackjack, where under some rules the player wins if he has any
> hand of five cards and does not bust (exceed 21). I figured if any
> message tripped 5 positive tests, the chances of it being non-spam were
> ve
Kingsley G. Morse Jr. wrote:
> Good point. Combinations of some rules may be more
> indicative of spam than others.
>
> It would be great if the GA could infer the boolean
> logic, as well as the scores.
It's possible that you could group the rules that matched, and feed it
into the score gener
On Thu:10:01, Matt Sergeant wrote:
> Kingsley G. Morse Jr. wrote:
> > Good point. Combinations of some rules may be more
> > indicative of spam than others.
> >
> > It would be great if the GA could infer the boolean
> > logic, as well as the scores.
>
> It's possible that you could group the ru
On Wednesday, May 29, 2002, at 10:51 PM, Michael Moncur wrote:
>> I came up with the name "Five-Card Charlie", which is a reference to
>> the
>> game of Blackjack, where under some rules the player wins if he has any
>> hand of five cards and does not bust (exceed 21). I figured if any
>> mess
On Thu, May 30, 2002 at 10:01:27AM +0100, Matt Sergeant wrote:
> Kingsley G. Morse Jr. wrote:
> >Good point. Combinations of some rules may be more
> >indicative of spam than others.
> >
> >It would be great if the GA could infer the boolean
> >logic, as well as the scores.
>
> It's possible that
Duncan Findlay wrote:
> On Thu, May 30, 2002 at 10:01:27AM +0100, Matt Sergeant wrote:
>
> Clearly, we can not do this with EVERY combination, unless Craig has a
> lot of CPU to spare. There are just under 400 rules right now. If we
> ended up with 400 tests, there would be 79800 doubles and 1058
Rob Winters wrote:
RW> SA does not give any credit to the cumulative effect that would be obvious
RW> to any human reading the "tests=" line, let alone the message itself. I
RW> mean, look at *this* one!"
This is currently true, and is basically a function of how the score-setting
(and score eva
It's now on again in CVS -- still need to rescore the phrases against the
updated corpus, but the scores in there now are decent.
C
rODbegbie wrote:
r> This reminds me... Whatever happened to the discussion of turning Spam
r> Phrases back on? I think Craig said it was something he'd be looking
Michael C. Berch wrote:
MCB> No, because the GA (if I understand how it is used correctly) only
MCB> considers rules individually, and not in combination (by number or
MCB> specifically). What I and some others have argued is that in many cases
MCB> tripping 5 low-scoring rules may be a better i
Duncan Findlay wrote:
DF> Clearly, we can not do this with EVERY combination, unless Craig has a
DF> lot of CPU to spare. There are just under 400 rules right now. If we
DF> ended up with 400 tests, there would be 79800 doubles and 10586800
DF> triplets.
We really don't care about *EVERY* combin
RW> How about a "bonus" for cumulative effect? Why not do a second-level
RW> analysis after scoring; something like:
RW>
RW> 3 positive score matches - add 1.0
RW> 4 positive score matches - add 2.0
RW> 5 positive score matches - add 4.0
RW> 6 positive score matches -
22 matches
Mail list logo