http://bugzilla.spamassassin.org/show_bug.cgi?id=4546





------- Additional Comments From [EMAIL PROTECTED]  2005-08-19 14:09 -------
Client/server user authentication is difficult, or at least it will take a lot
of thought and work to get right. Unless someone comes up with some standard way
of doing it that can be just dropped in to our code, it isn't something to add
for the 3.1 time frame. auth-ident is not meant to be a security measure except
under specialised circumstances. We could think about whether adding the ability
to use client side SSL certificates would make it possible to configure
spamc/spamd with secure user authentication without much change to the code, but
I think even that small change is too much for 3.1 and is not a sufficiently
complete solution.

On the other hand, as Duncan points out in the bug description, we have ended up
with something that has a vulnerability in any configuration that allows
untrusted people to have personal Bayes databases. I don't see why the problem
is any different with SQL vs DBM, as in both anyone can call spamc -L with any
-u argument that they want.

If we only document this, we would have to say that spamd can be used with
individual Bayes databases only in configurations in which identd authentication
can be trusted, or in which something like a VPN or ssh tunnel is used along
with accepting connections only from localhost. That seems to me to be too
restrictive.

I agree with Duncan that adding a spamd command line option that disables spamc
-L, plus adding documentation, is a sufficient step to take that is simple
enough to add in for the 3.1 ship. That would allow a sysadmin to run spamd on a
server, maintain individual Bayes databases for users, not have to count on
--auth-ident, and set up something ad hoc for users to train on errors such as
special IMAP folders or special email address to forward mistakes to.

We can then open an enhancement bugzilla ticket to consider how to get secure
spamc/spamd user authentication for 3.2.




------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

Reply via email to