Marc Perkel wrote: > > On 01/21/16 17:23, Michelle Sullivan wrote: >> Marc Perkel wrote: >>> Because it works on NOT matching instead of matching they don't get >>> the same advantages as matching systems. If they try to fake one of >>> those subjects it only helps them pass one of hundreds of tests. So at >>> best their fake will make an opportunity to detect ham become neutral >>> on the phrase. >>> >>> Poisoning doesn't work against my system. Misspelling makes it easier >>> to identify spammers. And eventually the spammer has to convince you >>> to do something and that's where they get caught. >>> >>> >> >> If you're doing it just on the subject, ok I'll go with that.. >> >> If you're doing it on the body you're probably working against prior art >> of a 2004 patent. >> >> Regards, >> >> Michelle >> > > OK - which patent is that? I'm doing it against the first part of the > body, the text in links, the name part of the from address, the php > script references, the header structure, and behavioral flags. > > Oh God only knows... but back in 2004, I (with 2 others) developed a system to detect real emails based solely on the body of the message, and to reject everything else as spam...
It was 99.996% efficient on *english* mailboxes (we only tested it against english mailboxes) ... I know a patent was applied for, I don't know if it was granted... If it wasn't granted it was because of prior art... if it was granted you may be using prior art granted to me an 2 others... from what I recall it may not have been granted because of about 20 patents that Microsoft were granted.... Either way looking for 'not spam' was something I suggested and developed way back in 2004... The 'SORBS Spam Firewall' was something that took the concept spamfilters and flipped it on its head with the thought... forget looking for spam because spammers will keep trying to evade filters, and look for real email ... because real email doesn't try to change/evolve and avoid... And here's a link to the PR... http://117.53.167.71/asia/radio/onairhighlights/revolutionary-spam-firewall and as a matter of interest (especially when re-reading what I said)... when dealing with companies that want to fund to make money and don't actually care about a product... forget timelines... Uniquest sat on it and killed it because they didn't want to make anything but a 'quick buck' (as a matter of interest i still have the code and still run it using the original trained set.. and it's still >87% correct (not bad at >10 years I think.) Regards, Michelle -- Michelle Sullivan http://www.mhix.org/ _______________________________________________ mailop mailing list mailop@mailop.org https://chilli.nosignal.org/cgi-bin/mailman/listinfo/mailop