[VulnWatch] Local Denial Of Service Attack Against Apple MacOS X, MacOS X Server, and Darwin.

From: Marukka (marukka_at_consoleconductor.com)
Date: 12/30/03

  • Next message: Peter Winter-Smith: "[VulnWatch] Re: NetObserve Security Bypass Vulnerability" 
    Date: Tue, 30 Dec 2003 16:46:15 -0600
To: <vulnwatch@vulnwatch.org>

    Advisory Name
    Local Denial Of Service Attack Against The SecurityServer Daemon In MacOS X,
    MacOS X Server, And Darwin.

    Release Date
    12-30-03

    Effected Platforms
    Apple MacOS X, MacOS X Server, and Darwin.

    Author
    Matt Burnett (marukka@consoleconductor.com)

    Vendor Status
    No patch has been released as of 12-30-03.

    Overview
    SecurityServer is a daemon used by the effected platforms to provide
    authentication, authorization, password dictionary (Keychain), and other
    services. It is possible for any user to cause the SecurityServer daemon to
    crash. When this happens it will have a cascading effect crashing other
    processes, and leaving the system in an unusable state.

    Details
    It is possible to cause the SecurityServer daemon to crash by unlocking a
    locked keychain and specifying a very long password. When the SecurityServer
    daemon crashes, it will have a cascading effect crashing other processes
    that rely on it. Since MacOS X, and many GUI and CLI programs rely on the
    authentication, authorization, and password dictionary services provided by
    SecurityServer, this can cause undefined behavior of those processes.
    Typical behavior for programs such as login, sshd, su, and sudo are that
    they will say you entered an invalid password. Typical behavior for most
    applications that use the authorization services is that they will hang. It
    is not possible to manually restart the SecurityServer daemon unless you
    already have a existing root shell open before the attack has taken place,
    since SecurityServer must be launched as root and it does no have the suid
    bit enabled. Even when it is relaunched after an attack has taken place, it
    will leave the system in an unusable state. Logins via the GUI (loginwindow)
    will not work and authorization services will not work either. The only
    realistic way to recover from an attack is to reboot the machine. I have not
    fully investigated the extent that this attack could be exploited, or its
    effect on a system as a whole since so many programs and applications rely
    on the SecurityServer daemon.

    Proof Of Concept Code
    To build this code run ³gcc <file name> -framework Security ­o
    CrashSecurityServer²

    #include <Security/Security.h>
    int main(int argc, const char *argv[])
    {
        SecKeychainRef defaultKeychain;
        SecKeychainCopyDefault(&defaultKeychain);
        SecKeychainLock(defaultKeychain);
        SecKeychainUnlock(defaultKeychain, 0xFFFFFFFF, "password", true);
        return 0;
    }

    Vendor Response
    Apple Developer Connection told me that Apple does not give release dates
    for patches.

    Timeline ??-??-?? Flaw discovered. Most likely in MacOS X Public Beta
                            or 10.0. I do not remember the exact date.
                11-20-03 Vendor is notified of flaw and is supplied with
                            proof of concept code.
                12-29-03 Asked vendor for status update. Apple Product
                            Security referred me to Apple Developer Connection.
                            Apple Developer Connection informed me that Apple
                            does not give release dates for patches.
                12-30-03 Advisory and proof of concept code
                            released.

    Recommendation
    As of the release of this advisory Apple has not yet released a patch. I am
    unable to release a patch because Apple only makes portions of the code
    needed to build SecurityServer available to the public. My only
    recommendation is for you to allow only people your personally trust onto
    your systems if they are effected by this flaw.

  • Next message: Peter Winter-Smith: "[VulnWatch] Re: NetObserve Security Bypass Vulnerability"