From: ph@netcom.com (Peter Hendrickson)
Date: Thu, 19 Dec 1996 21:18:19 -0800 (PST)
To: "Timothy C. May" <cypherpunks@toad.com
Subject: Re: Executing Encrypted Code
Message-ID: <v02140b03aedfd5494638@[192.0.2.1]>
MIME-Version: 1.0
Content-Type: text/plain


At 9:20 PM 12/19/1996, Timothy C. May wrote:
>> However, why not use "beacons"?  The clock could have a built-in timer
>> that needs to be reset once a month from an authenticated source.  This
>> assumes the presence of net connectivity, but that's not a terrible
>> assumption.

> I mentioned "beacons" in the portion of my message you did not quote here.

Gack!

> As for why they are not being used, they don't exist.

Here's how I would do it.  When the processor wants to update its
clock, it generates a random number and encrypts it for the trusted
time source.  The trusted time source decrypts its message to get
the random number.  It timestamps it, encrypts it, and sends it
back.

This means you can't replay old time messages to keep using your
old software.

Is it possible to have a little clock and rechargeable battery on
a chip?  If so, then this technique should be easy to use.

If not, then the processor can count the number of cycles it runs and
use that as an approximate means of deciding when to check the time.

Or, it could demand a time update every time it is power cycled.

Peter Hendrickson
ph@netcom.com