On Fri, 16 Jul 2010 14:13:20 +0200, Stevan Bajić <ste...@bajic.ch> wrote: > On Fri, 16 Jul 2010 13:32:53 +0200 > Cyril' <ds...@minou.biz> wrote: > >> Hi all, >> > Hello, > > >> Let's assume I have an organization where I want an antispam installed, >> but the users are too dumb/busy/unhelpish to process their own receied >> spam and forward it to the retrain process, in any way. (Not an academic >> problem, it's the story of my life). I don't even talk about using a web >> gui for that. >> Would dspam be the right tool for them?? >> > Yes. > >> I can handle doing the retrain >> myself for the first days/weeks, but after that, each new false >> negative, not retrained, will let dspam believe more and more that it is >> ham, and at the end, nothing more will be filtered, right? >> > Wrong. If you use TOE ([T]rain [O]n [E]rror) then a not corrected error > (either False Postive or False Negative) will not influence the tokens at > all. The only influence it will have is that the counter for each class > (Spam / Ham) will be increased and that counter will have a influence on > the accuracy but in my experience the wrong classification will > balance/compensate each other. So at the end your tokens will not get > better nor will they get worse. They +/- stay on that level and the > accuracy will do the same. > > >> Or could there be a strategy to accomodate this? >> > YES! Use the group support of DSPAM. That will help to boost accuracy. For > example using a merged group will help to boost the accuracy for all group > members at once without the need to train each user individually. Or if you > have groups of users sharing +/- the same type of mail content then put > them in a shared or shared,managed group and correct errors there from time > to time. > >> (No, changing this customer is not an option for my business...) >> > What you have there is something most of us have to deal with. It is > pretty common to have users not being able or not being capable to train. > >> Ideas welcome, >> > What MTA are you using? Maybe I could help to harden your setup? I have so > far helped two users from the DSPAM mailing list to harden their setup and > I would say that those little things have helped to cut their Spam rate by > factors. > > @Marko Weber and Paul Cockings: If you are reading this... could you post > your experience with the hardened setup? I think Cyril could benefit from > some input.
My Experience with the hardened setup is very ver very good. i get so less spam in compare to before, its awesome. The Tools and Setup "before" dspam do great jobs, so dspam has not much to do. only from time to time when a spam gets thru, i train dspam with dovecot-antispam, the user just moves the wrong identified spam or not correct detected spam "from" or "to" folder SPAM. When i get 5-10 Spam now a month, thats much. The hardened Setup really safes me alot time and nerves and that of 2 little cutsomers. > > >> Cyril' >> > -- > Kind Regards from Switzerland, > > Stevan Bajić > > ------------------------------------------------------------------------------ > This SF.net email is sponsored by Sprint > What will you do first with EVO, the first 4G phone? > Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first > _______________________________________________ > Dspam-user mailing list > Dspam-user@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/dspam-user ------------------------------------------------------------------------------ This SF.net email is sponsored by Sprint What will you do first with EVO, the first 4G phone? Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first _______________________________________________ Dspam-user mailing list Dspam-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/dspam-user
