Markus Braun wrote: > > At the moment i use spamassassin for spam protection. > Also Autobayes is activated. > > I know when autobayes make some mistakes with ham or spam, that i can > correct it manually. > > e.g. > sa -- learn (the file) > > But what does sa--learn keep in mind? > > On the next day, a spam emails als come from another email adress. > But the adressname and the header information are the same. > > So my question is, how can i make it that these emails are also > marked als spam.
A couple of things to keep in mind about Bayes... First, it doesn't just look at email addresses. Bayes processes the entire file. It operates by keeping statistics on how many hams and spams certain words appear in. Then when it evaluates a messages, it takes all of that into consideration to determine the probability that this new email is spam. So it doesn't matter that spammers keep changing email addresses. As long as the content is similar, Bayes will score it about the same. Second, sa-learn will adjust the database, but don't expect a major change immediately. The Bayes database changes gradually as it learns mail. You may have to learn four or five of a new type of spam before the database adjusts enough to start calling it spam. On the other hand, if a new spam run is similar to an old one, Bayes may pick it up immediately without any further learning. This is one of the advantages of using Bayes. Also keep in mind that Bayes is just one component of SpamAssassin. Just because SA marked a message as spam, doesn't necessarily say anything about the Bayes score. However, if you consistently get spam with a low Bayes score (BAYES_00 in particular) you should seriously consider wiping the database and starting over. Third, Bayes knows which messages it has already learned, so you can't mess things up by learning a message more than once. If you learn it once as ham and then learn it again as spam, the last run will override the previous one. -- Bowie
