> >>> i'm trying to teach my SA whats spam > >>> > >>> it's a brand new out of box SA, i have few domains that i dont get > >>> anything but a spam and on the top seems like from same spamers as > >>> they "picked" emails that they thought would be good to spam and keep > >>> on spaming them
Different domains -- are these different users, too? Do you have a site-wide Bayes setup? The training and scanning user must be the same. Did you train as the scanning user? > >>> yet, when more of some what same email comes in it still can't > >>> determinate if its spam or not... I assume you do have Bayes enabled, and that the training user is the same as the scanning user. Are you positive the FNs are due to Bayes? You didn't show any evidence. Which rules do these messages trigger? If need be, just upload a raw sample including all headers and body somewhere, your own webspace or a pastebin, and provide the link. > > i don't remember how but last time i was able to pull some sort of > > stats and it had plenty of ham emails as well Yup, sa-learn --dump magic. ;) > 0.000 0 3 0 non-token data: bayes db version > 0.000 0 5603 0 non-token data: nspam > 0.000 0 1066 0 non-token data: nham > 0.000 0 146370 0 non-token data: ntokens That's sufficient for Bayes to kick in, with the default thresholds of 200 messages each. Did you gather these stats -- and do the manual training -- as the *same* user that scans your incoming mail? -- char *t="\10pse\0r\0dtu...@ghno\x4e\xc8\x79\xf4\xab\x51\x8a\x10\xf4\xf4\xc4"; main(){ char h,m=h=*t++,*x=t+2*h,c,i,l=*x,s=0; for (i=0;i<l;i++){ i%8? c<<=1: (c=*++x); c&128 && (s+=h); if (!(h>>=1)||!t[s+h]){ putchar(t[s]);h=m;s=0; }}}