13.12.2012 10:35, Nikolaos Milas kirjoitti: > Hello, > > We have a Postfix 2.9.4 with Amavisd-new 2.8.0/ClamAV/Spamassassin on > Centos 6.3 (x86_64) used as an incoming mail filtering gateway which > relays clean (only) mail to an internal final destination mail server > (Postfix/Dovecot). > > We are not using Maia or other such systems supporting per user > filtering/spam rule settings. (We want to keep the setup as simple as > possible and controllable by the admins.) > > I would like to setup a spam mailbox (like: [email protected]) where > users can submit what they deem as spam, and which can be used to > train SpamAssassin. The idea is to receive reported spam as attachment > (one per mail), so we can process it later, e.g. by saving spam mail > files in a directory and running: > > sa-learn --spam /path/to/spam/folder > > Questions: > > * Do regular eml files (of spam mail messages) qualify for such use > (i.e. does sa-learn understand them properly)? > * What happens with spam messages forwarded from Outlook clients which > will forward as msg and not as eml? Do we need a msg-to-eml > converter script to process these files? > > Has anyone here set up something like that and could suggest/provide a > script or similar, which would accomplish this task (I don't want to > re-discover the wheel)? > > Any advice will be appreciated. > > Thanks in advance, > Nick > >
The best solution might be an IMAP based mail server in the final server for the Users. Each and every User should have a special folder for the purpose. Then they could simply drag the Spam-to-train and possibly Ham-to-train to the folders. Separate Cron job then could process the samples and train the bayes. --jarif -- Your mode of life will be changed for the better because of new developments.
signature.asc
Description: OpenPGP digital signature
