[cap-talk] password capabilities & object capability model?

[Tony Bartoletti]

At 07:18 PM 8/21/2008 -0700, MarkM wrote:

>But the first part of the protocol of "tell me the secret number and I'll 
>give you the money" is comparison of large unguessable numbers. To me, 
>this is the simplest degenerate case of crypto.

Technically, you are right.  "Crypto" (from Greek "kruptos") simply means 
hidden or secret, and surely a swiss number must remain secret to maintain 
its utility...

Many people use the term "cryptographically strong" to describe (say) the 
MD5 or SHA1 hash of a bit-string, even though all operations are "in the 
clear" and nothing is being "encrypted" - the idea being that the precursor 
bit-string (or any hash-wise equivalent) is essentially unguessable, given 
only the hash.

Problem is, most folk who hear "crypto" think of "cyphers", or 
cryptographic *translations*.

Aside:  I recall (circa 1996, SPKI list) Ron Rivest introduced the 
definition-busting "chaffing and winnowing", to demonstrate that a strong 
digital signature capability can be employed to effect strong "encryption", 
even when everything is "in the clear":  Alice sends a message M to Bob, 
placing each individual bit of M in its own "signed packet", a 3-tuple 
(seqno,bit,sig(seqno,bit)).  Alice, or even Fred downstream, can read the 
flow and add to that flow a complement to each packet, retaining the seqno, 
flipping the bit, and placing garbage in the sig-part.  Only those who know 
to perform the corresponding signature verification know whether to accept 
the 0 or 1 bit in each case.

Efficiency is another matter ...

Cheers!  ____tony____