On 28 Jan 2011, at 18:39, Lawrence @ Rogers wrote: > On 28/01/2011 2:53 PM, David F. Skoll wrote: >> On Fri, 28 Jan 2011 18:10:08 +0000 >> Dominic Benson<domi...@lenny.cus.org> wrote: >> >>> Recently, in order to balance the ham/spam ratio given to sa-learn, I >>> have started to pass mail submitted by authenticated users to >>> sa-learn --ham. >>> I haven't seen any mention of this strategy on-list or on the web, so >>> I'm interested in whether (a) anyone else does this, and (b) is there >>> a good reason not to do it that I haven't thought of? >> It's possibly a good idea, but you want to be really careful of one >> thing: Make sure your users are savvy enough not to have their >> accounts phished. It'll take just one compromised account that blasts >> out a spam run to destroy the usefulness of your Bayes data. >> >> Regards, >> >> David. >> > Agreed. I was considering the same idea at one point, and came to the same > result. One person could poison the DB completely.
Granted. I'm not hugely worried about it in my setup; there are only 60-70 users and they're pretty well trained to just e-mail me if there is anything unusual. Worst case, restore the bayes DB from backups. In a bigger / broader set-up, I imagined doing the same thing but with per-user Bayes: then you'd just clear any affected user's database, without making a mess of everyone else's data.