Re: Security Problem in MH 6.8.4

From: Cy Schubert - ITSD Open Systems Group (cschuberat_private)
Date: Tue Jan 20 1998 - 00:20:44 PST

  • Next message: Prince Ctrl: "Re: Security Problem in MH 6.8.4"

    >   Description:
    >       Due to lack of security checks there is a standard stack smashing probl
    em.
    > Local user can execute code as root.
    >
    >     Let's see.
    >
    > [tascon@archivald]$ id
    > uid=500(tascon) gid=500(tascon) groups=500(tascon),100(users)
    > [tascon@archivald]$ cat /etc/redhat-release
    > release 5.0 (Hurricane)
    > [tascon@archivald]$ ls -l /usr/bin/mh/inc
    > -rwsr-sr-x   1 root     mail        82972 Oct 15 18:06 /usr/bin/mh/inc
    > [tascon@archivald]$ /usr/bin/mh/inc
    > inc: no mail to incorporate
    > [tascon@archivald]$ /usr/bin/mh/inc -host XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
    XXX[...]
    > XXXXX      <---- (2000 X's here)
    > Segmentation fault
    >
    > ^^^^^^^^^^^^^^^^^^   Dangerous isn't it?
    
    Under FreeBSD (while using NIS), specifying an argument greater than 1024
    bytes causes an infinite loop with the following message being printed;
    
            yp_match: clnt_call: RPC: Can't encode arguments
    
    Under FreeBSD (while not using NIS) and DEC UNIX (using MH 6.8.4 instead of
    the MH 6.7 that is supplied with DUNIX), specifying an argument greater than
    4600 bytes, the error is handled properly and the the following message is
    produced;
    
            inc: no servers available
    
    Without having a chance to recompile MH with -g and testing it under gdb, I
    suspect that the Linux segmentation violation and the FreeBSD NIS loop are
    occuring within the gethostbyname() call or some libc call made by
    gethostbyname().
    
    One may argue that the FreeBSD NIS loop may constitue a DoS attack as it chews
    up a fair amount of CPU time, however there are probably better ways to bring
    a system to its knees.  I'd characterize the FreeBSD NIS bug as more of an
    annoyance.
    
    The Linux bug is definitely a security issue.  I suspect it to be in RedHat's
    implementation of libc.
    
    >
    >    Local exploit exists for that option. Note that MH isn't even configured.
    > It's as the installation of RedHat 5.0 left it. Note also that MH is intalled
    > by deffect with RedHat 5.0.
    >
    > Solution: Uninstall this package or remove the suid-bit until patch becomes
    >           available.
    
    Another solution might be to recompile MH without POP support, however that is
    just a band-aid solution and doesn't really fix the problem with RedHat's libc.
    
    >
    > MH also installs another suid-program: msgchk. It's also posible to get a
    > Segmentation fault whith the same option, but I haven't been able to exploit
    > it. I have worked on it quite a few. Could someone probe it a little deeper??
    >
    >   Greetings
    >
    >
    > ----o-------------------------------o-------------------------------------o--
    --
    >   Space reserved to describe      /          Cesar Tascon Alvarez
    >     my job when I got one.      /       University of Valladolid (SPAIN)
    >  Yes, I'm just a student ;)   /               tasconat_private
    > ----o-----------------------o---------------------------------------------o--
    --
    >
    
    
    
    Regards,                       Phone:  (250)387-8437
    Cy Schubert                      Fax:  (250)387-5766
    UNIX Support                   OV/VM:  BCSC02(CSCHUBER)
    ITSD                          BITNET:  CSCHUBERat_private
    Government of BC            Internet:  cschuberat_private
                                           Cy.Schubertat_private
    
                    "Quit spooling around, JES do it."
    



    This archive was generated by hypermail 2b30 : Fri Apr 13 2001 - 13:40:01 PDT