RE: [Symantec Security Advisor] Symantec Security Check ActiveX Buffer Overflow

[Chris Wysopal <weld () vulnwatch org>]

On Tue, 24 Jun 2003, Jason Coombs wrote:

> 1) Does this ActiveX control bear a digital signature? If so, the problem it
> causes does not go away simply because there is a new version available from
> Symantec. An attacker in possession of the bad code with its attached digital
> signature can fool a victim whose computer does not currently have the
> vulnerable code installed into trusting the ActiveX control due to the fact
> that Symantec's digital signature will validate against the trusted root CA
> certificate present by default in Windows -- the existence of the digital
> signature on the bad code effectively transfers ownership of millions of other
> people's computers to anyone who should become interested in attacking those
> computers; it is extremely important that Symantec take further action above
> and beyond compiling a new version of the affected code because of the ongoing
> threat posed for the duration of the validity of the digital signature.

You are absolutely right about attackers using the old control to carry out
an attack.

The new control should have a new CLSID and the kill bit should be set for
the old control's CLSID.  Information from the Microsoft knowledge base on
how to set the kill bit is here:

http://support.microsoft.com/default.aspx?scid=http://support.microsoft.com:80/support/kb/articles/q240/7/97.asp&NoWebContent=1

Unfortunately the only way to get this kill bit to be set on the majority
of machines is to get Microsoft to do it through a Windows update. Until
that happens the old signed control can be used by attackers.

This is the real flaw in the system.  The kill bit is only useful to
Microsoft as Symantec has no way of getting all Windows users to set this
bit on the bad CLSID before they are attacked.  Perhaps Microsoft should
allow other vendors to send them CLSIDs to kill.  Or maybe they already do
allow this but it is not publicized.

-Chris


> Sincerely,
>
> Jason Coombs
> jasonc () science org