On Mon, 1 Jan 1996, amp wrote:

DS> I'd think you could have the server safely accept # N, N-60 sec, and DS> N+60 seconds; and adjust the server's idea of your card's clock speed DS> from that.

DS> What new risk would that create?

i would figure the server would give a minute or so for slippage. basically the risk is that it would give someone 3 minutes to do a brute force attack rather than one. if you have decent security on the server side, i.e., disallow the card for 5 minutes or more after 3 or so failed attempts, brute attacks would be minimized. however, if the actual window for a single code is 3 minutes, that increases your chance of hitting it as 3 separate numbers would be valid for a given card at any given time.

START <attila> Bank wire systems over the SWIFT private wire are time synched much closer than a minute although I have never been given more of an answer than that. given that you have a tolerable high speed link, and are not dealing with an overloaded concentrator at the telco -> carrier inferface or an overloaded server, I believe you can solve most of the windowing problem by: 1. client sends number and time to server 2. server send what it thinks as time to client 3. client can place a delta on servers time for local time 4. enter PIN, etc. and you are working with a much narrower window. the security risk does not appear to increase from the exchange times and entering the PIN and letting the normal progression go forward once v. just monitoring a series of successive verifications trying to effect a pattern in the hash. Secure-ID seems to be a one-time time-based single use pad; to me, using a time exchange initiator has the advantage of a smaller window, and fewer problems with client machines running on strange times which require sloppier time windows. END <attila>