Re: Security of Secret Algorithm encruption

"William L. Bahn" <william@toomuchspam.net> writes:
> What I'm curious about is, under those conditions, how complicated
> the algorithm would need to be in order to have a good chance at
> defeating a concerted effort to break the cipher. In other words,
> how difficult is it to attack an arbitrary and unknown algorithm?
> What kind of attacks would be used?

secret algorithms can be considered from the standpoint of
defense-in-depth ... where effort for attackers is additive based on
the depth of defense and assuming straight-forward frontal attack.

the counter argument is multiple security layers tend to add
complexity ... making it more dfficult for the defenders to correctly
maintain (or even understand) at all times (KISS) and even creating
cracks that attackers can use for compromise (again KISS).

there was a story earlier about some encryption device that used
secret algorithm that was supposed to be widely deployed ... however
the units were classified and the local security officers had to sign
away their future; so the security officers kept the units locked in
vaults with extremely limited access. as a result the units weren't
being used for their intended purpose (in part because the loss of a
unit was perceived as being an enormous threat).

a corollary is that there tends to be scale-up and long-term
deployment issues with secret-based operations. if nothing else, for
infrastructures dependent on secrets, it is convenient that secrets
are easily changed in the face of possible compromise ... secret
algorithms tend to be harder to replace/update than keys when there
has been compromise.

so the threat models are not only how difficult are frontal attacks
... but also given some compromise by any method ... how difficult is
the remediation.

sometimes strict technology orientation is myopic and focused on
single point events. long term institutional issues frequently have to
also consider long term (and continued) operational characteristics
spanning a large number of possible different events.

and at least for human element and operational characteristics
... KISS.

long ago and far away ... there once was a corporate rulling that
passwords had to be of certain complexity and changed on frequent
intervals. a parady was done ... minor reference.
http://www.garlic.com/~lynn/2001d.html#51 A beautiful morning in AFM.
http://www.garlic.com/~lynn/2001d.html#52 A beautiful morning in AFM.

as a separate issue, i asserted that key-based locks were more
vulnerable to brute-force attacks than passwords ... and if frequency
of change was a valid countermeasure ... then all physical locks
should be rekeyed every couple hrs ... and new physical keys issued
for everybody.

-- 
Anne & Lynn Wheeler | http://www.garlic.com/~lynn/