A bit delayed in answering, sorry I've been ignoring [SAtalk] for a few days to
get some Real Work done. Matt, I agree the porn rules are pretty lame
currently. They need revamping. I've got a bug in bugzilla (#46) which was
reminding me to work on this, though it doesn't include the full details of all
the changes I wanted to make. Your suggestion of extending PORN3 further is
good, but I think it's critical to break it up somewhat so that "wild adult
asian ferrets" isn't lumped in the same scoring bucket as "teen cum fuck".
It would be helpful if you make a summary of your ideas posted to the list
here and append it to bugzilla#46
Thanks,
C
Matthew Cline wrote:
> Date: Wed, 13 Mar 2002 02:57:53 -0800
> From: Matthew Cline <[EMAIL PROTECTED]>
> To: [EMAIL PROTECTED]
> Subject: [SAtalk] PORN ideas 2
>
> Ooops, hehe; sorry about that...
>
> Interesting subject for a technical mailing list, huh? :-)
>
> I started fiddling around with the PORN_3 rule because it wasn't catching any of the
>porn spam that I got.
>
> body PORN_3
>/(?:(?:\bcum|\borg[iy]|\bwild|fuck|\bteen|\baction\b|spunk|\bpussy\b|\bpussies\b|suck\b|sucking\b|\bhot\b|\bhottest\b|\bvoyeur|\ble[sz]b(?:ian|o)|\banal\b|\binterracial|\basian\b|\bamateur|\bsex+\b|\bslut|explicit|xxx[^x]|\blive\b|celebrity|\blick|\bsuck|\bdorm\b|webcam|\bass\b|\bschoolgirl|\bstrip|\bhorny\b|\bhorniest\b|\berotic|\boral\b|\bpenis\b|\bhard-?core\b|\bblow[
>
>-]*job|\bnast(?:y|iest)\b|\bporn|\bwhore|\bsexfest|\bnaked|\bnude|\bvirgin|\bnaughty\b|\bnaughtiest\b|\bgirls\b|\bcelebs?\b|\bbabes\b|\badult\b).{0,15}){3,}/i
>
> I got rid of the last "\b" "\bschoolgirl\b", because that prevented it from matching
>"schoolgirls", and I added a "-?" to the middle of "hardcore", because it's
>ocasionaly written "hard-core". Then I added a lot more words. They are (excluding
>the "\b"s):
>
> - whore
> - sexfest
> - naked
> - nude
> - virgin
> - naughty
> - naughtiest
> - girls
> - celebs?
> - babes
> - adult
>
> However, that still only caught a few of the porn spams I got. I figured that we
>could add some more sensitive rules that would catch things PORN_3 wouldn't catch,
>but would also detect anything that PORN_3 would; the points assigned to PORN_3 would
>automatically be redistributed out among the different porn rules by the GA. In
>fact, we could make several layers of porn rules, from very sensitive, to slightly
>sensitive, to PORN_3 sensitive, and as spam trigger less and less sensitive rules,
>the points would add up.
>
> Here's what I came up with:
>
> ---------------
>
> # Words which will almost never be seen outside of porn spam
> body PORN_9 /(?:sexfest|cumshot)/i
> describe PORN_9 Uses words and phrases which indicate porn (9)
>
> # Words which mainly show up in porn spam
> body PORN_10
>/(?:\bslut|\bwhore|\bxxx\b|\bporn|\b(?:Asian|Japanese|oriental)\s+(?:girls|schoolgirls)\b|\bbabes\b|gang[
> -]?bang)/i
> describe PORN_10 Uses words and phrases which indicate porn (10)
>
> # Words which often show up in spam, but also show up somewhat often in
> # non-spam
> body PORN_11 /hard-?core/i
> describe PORN_11 Uses words and phrases which indicate porn (11)
>
> # Like PORN_3, but it only needs to match two terms instead of three. To
> # compensate for only matching two terms, so as to not generate too many
> # false positives, there can only be up to seven characters between terms,
> # rather than the 15 for PORN_3; it also uses a short list of terms.
> body PORN_12
>/(?:(?:\bxxx|\bsex|\bslut|\bwhore|\bhottest\b|hard-?core|\bhorny\b|\bhorniest\b|\bvirgin|\bnaughty\b|\bnaughtiest\b|\bwebcam||\ble[sz]b(?:ian|o)
> describe PORN_12 Uses words and phrases which indicate porn (12)
>
> -------
>
> Oh, and I also added "lust", "panty" and "panties" to the PORN_4 rule:
>
> uri PORN_4
>/^https?:\/\/[\w\.]*(?:xxx|sex|anal|slut|pussy|cum|nympho|suck|porn|hardcore|taboo|whore|voyeur|lesbian|gurlpages|naughty|lolita|teen|schoolgirl|kooloffer|erotic|lust|panty|panties)\w*\./
>
>
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
