[sci.crypt] authentication without encryption

karn@faline.UUCP (Phil R. Karn) (11/21/86)

I am interested in "systematic" authentication schemes, i.e., methods
for verifying the authenticity of a message without hiding the meaning
of the message itself.  This is useful for certain special situations
such as amateur radio, where encryption per se is illegal but methods to
prevent unauthorized use of a remote radio control link are OK.  I seem
to remember hearing that the banking industry has some ANSI standards
for this purpose, but I haven't been able to find copies.

I have one idea for this that I'd like comment on. Basically, you
encrypt the message with DES in the cipher feedback mode. Then you
transmit the final 64 bits of the ciphertext after the original
plaintext message. This adds a a 64-bit "crypto checksum" which guards
against attempts at spoofing or modification by someone who doesn't know
the DES key.  I don't see any way that the message could be modified
(e.g., by swapping words or bytes of an intercepted message) without
perturbing the crypto checksum in some totally unpredictable way. About
the only thing the attacker could do is to play back old messages. If
your network is based on datagrams, though, this isn't a problem. All
end-to-end protocols on datagram networks are already designed to filter
out duplicate packets, since datagram networks occasionally duplicate
packets for reasons other than malicious spoofing.

Comments?

Phil