On Sun, Jul 02, 2000 at 08:44:23PM -0700, Brian D. Winters wrote:
> Initially I thought I saw your point, but I was wrong.  You don't seem
> to be making any sense.
> 
> On Sun, Jul 02, 2000 at 10:17:23PM -0400, Adam McKenna wrote:
> [this sentence originally came after the next quoted block]
> > If he can find a security hole that allows him to read files
> > that don't belong to him, he now has the entire list of passwords.
> 
> Make the list readable only by root.  Now a local user effectively
> needs root access to read the APOP secrets.  Once that local user has
> rooted the box, I don't see why it matters that the secrets were
> cleartext.

There have been several exploits in the past that allowed a local user to
read files on the system without obtaining root.  Granted, if someone found
a vulnerability like this he could read the shadow file, but at least the
passwords in the shadow file are encrypted (as opposed to the passwords in
the apop.secrets file).

> > That was entirely my point.  IMO the "security cost" of saving cleartext
> > passwords on the server is not worth the "security gain" of having POP3
> > passwords encrypted when the user checks his mail.  If someone is sniffing
> > pop3 passwords then he has the ability to (most likely) only obtain a small
> > number of passwords that way, as opposed to the attacker who has an account
> > on the server.
> 
> So you don't care if anyone with network access has "a small number of
> passwords"?  Why is one user password better than another?  If there
> are local root vulnerabilities present on the system, any single user
> account should be good enough to exploit them.  Allowing someone to
> sniff any number of passwords sounds like a Bad Thing(tm).
> 
> I have yet to work in an environment where it is harder to run a
> packet sniffer than it is to find a local root vulnerability.

As you've pointed out, an attacker with sniffing ability can already read the 
e-mail, which is the only thing that the password protects.  If a sniffer can
read the e-mail, then who cares if he has the password?  I'd rather make sure
that if he DOES get the password, it will be useless except for reading the
e-mail.

I guess the point I'm trying to make is that APOP is akin to putting all of
your eggs in one basket.  I believe in the security practice that passwords
should _never_ be stored in cleartext, and I really don't see a reason to
go against that practice, especially when there are other, better methods of
securing e-mail.

Also, in the day of $150 500-MHz cpu's, I'm not really convinced by the
system resources argument.

> > If you're concerned about email security, APOP is not worth it.  Go with SSL 
> > or another security model (like having virtual POP3 accounts that aren't UNIX
> > users).
> 
> I think the point you are missing is that APOP effectively creates
[...]
> overkill in a lot of situations.

You are entitled to your opinion, and I certainly won't stop you from using
APOP on any of your servers.  I am merely stating that I'll never use it on
any of mine.

> Brian
> 
> PS Somewhere else in this thread someone mentioned that the only APOP
> client they were aware of is Eudora.  FWIW, fetchmail also supports
> APOP.

Heh.  I'll leave my opinion of fetchmail out of this.

--Adam

Reply via email to