Re: Red Hat 6.2's ftp segmentation fault

[11a () GMX NET (Bluefish)]

> Try 'strings core'. You'll find your password nicely on display in
> plaintext. This is hard to exploit, but still poor security. (Although
> your password did just go across the network in plaintext, so a core
> dump is the least of your problems).

I assume the correct limits, or correct umask will solve that problem.

> The issue here is that your actual, cleartext password need *never*
> appear on any disk, anywhere at any time. If it's being stored or
> transmitted, it should be hashed or encrypted.

Obviously it has to be in memory SOME time, in order to encrypt (hash) it.
Remember that cores aren't created deliberatly!!

> The authentication client (e.g. ftp, ssh) should mlock(2) the buffer
> allocated for the password, and then memset(3) it as soon the
> authetication is complete. Non-root processes can't ordinarily lock
> memory, so you might find your password in swap. But there is no reason
> to hang on to the password once authentication completes.

Agree, that's a good idea. Of course, a badly encoded crypt() etc could
cause you some trouble ;)

..:::::::::::::::::::::::::::::::::::::::::::::::::..
     http://www.11a.nu || http://bluefish.11a.nu
    eleventh alliance development & security team