🚀 go-pugleaf

mark@hpcvxmsr.cv.hp.com (Mark Rowe) writes:

>I agree, it is not always clear cut, but I think the wrong questions are being
>asked.  It is not, "Is the use of XXX a covert channel?", but rather, "Was XXX
>intended to transmit information about YYY?".  It is (as you said) this intent
>that is often not clear cut.  That in itself is not a fault in the definition.
>It is often much easier to define what something is than it is to accurately
>identify one when it is encountered.

Okay we're now talking about several different issues.  First, there may
be parts of a system that handle multi-level data.  These parts are the
TRUSTED parts (Trusted Computing Base TCB) -- these are the only parts that 
can do things like transmit XXX about YYY.  I assume that you mean mean YYY 
is at a different classification level than XXX.  This may be an overt 
channel (intentional and part of the design) or a covert channel.

Overt channels should only occur in the TCB.  These are normal in any system
design...

There are other parts of a system that are UNTRUSTED.  These parts are
not intended to handle multi-level data.  Any time these parts
do something like transmit XXX about YYY it's a covert channel.  That is,
the designers didn't intend for this part of the system to do that. 

The original definition of a covert channel from the Nutshell book 
is a good one.  Adding fuzzy cases, and talk about MAC, DAC, etc only
confuse the issue here.  A COVERT CHANNEL is a just a communication
channel that VIOLATES the SECURITY POLICY.  If it is overt it will be part
of the security policy of the system.  That is, it will be part of the
functions defined for the TCB.


--f