Bugtraq mailing list archives

Re: a few bugs ...


From: lcamtuf () DIONE IDS PL (Michal Zalewski)
Date: Fri, 17 Mar 2000 10:07:45 +0100


On Mon, 13 Mar 2000, Maurycy Prodeus wrote:

1. In "Lotus Notes POP 1.0X" on NT platform. I'm not really sure ...
if you send a very long username ( about 2kb ) it disconnects without
any message. So it looks like classic buffer overflow :) I don't have
enough time to check it ( to download this packet :) )

Have you noticed GPF popup or BSOD on this Windows box? Anyone may confirm
this?

2. Mail agent programs like: standard ;P 'mail' from Berkeley
Distribution or mutt, elm perhaps other :), use sendmail arguments to
put email adress where luser wants to send mail. It's similar problem
to crontab's or lpd's bugs. Example: if you put line with Reply-To: -X
/dev/hda1 ;P or something like that :> to mail message and luser ( in
this case root ) stupid pushes OK,OK,OK :) ( ofz he'd want to reply )
it may write/destroy file ( /dev/hda1 :] ). I know it isn't good
example but I only wanted to show idea...

AFAIK most of mailers calls Sendmail with -bs (start in SMTP emulation
mode) or -t (scan headers for recipients) parameter. This is true eg. for
pine, mutt seems to be not vulnerable as well. Linux /bin/mail seems to
behave correctly, as well.

3. ntalkd from redhat distri or debian... in old version ( <=5.2rh and
<=2.0db) (I don't want to be wrong so I will not write it's version -
aleph bounced;P sic! ) it's known and patched but there wasn't
official post and it may be dangerous. There is fprintf() without
format. Another hard to exploit bug :)

Aham. According to ChangeLog:

26-Nov-1998:
        Fixed bug: the talkd announce message is passed as the format
          string to fprintf, so if it has %'s in it, we probably crash.

Announce message (assembled in ntalkd/announce.c) contains remote username
and remote hostname information, as well as some hardcoded texts like
"Talk request from...". Take a note - we're talking about fprintf, so,
assuming there's no interesting data in daemon address space (I don't
think so - it is not performing any authorization, etc, only reads utmp
entries), I don't think it might lead to anything except crash. And, as
it's started from inetd, I don't think it might have any security
implications ;)

Btw. Aleph, some time ago I described proftpd crash problem with LIST
parameter. Instead of playing with FUD, I've done some debugging and
realized it won't be _probably_ exploitable. As the result, you bounced
this post, but approved this one - for sure overFUDed ;>

_______________________________________________________
Michal Zalewski * [lcamtuf () ags pl] <=> [AGS WAN SYSADM]
[dione.ids.pl SYSADM] <-> [http://lcamtuf.na.export.pl]
[+48 22 551 45 93] [+48 603 110 160] bash$ :(){ :|:&};:
=-----=> God is real, unless declared integer. <=-----=


Current thread: