[Cryptography] Best internet crypto clock

Jerry Leichter leichter at lrw.com
Thu Oct 23 07:27:51 EDT 2014

On Oct 22, 2014, at 10:35 AM, Henry Baker <hbaker1 at pipeline.com> wrote:
> I like this AC hum idea for a crypto clock, except that:
> 1.  It is highly local, so you need recordings from your local power provider to provide a time base....
Somehow we went from using the AC hum as a forensic mechanism to using it as a clock.

The use of the 60 (or 50) Hz baseline power frequency to produce accurate electric clocks goes way back.  In fact, this was a usage specifically supported by the power companies:  While all the generators in a system need to be synchronized, there's no need for them to maintain long-range stability or remain centered at 60Hz.  But cheap, accurate, synchronized electric clocks were an early selling point, so the systems were built to actually stay close to the nominal center frequency, and were deliberately manipulated to long-term average stability.  To do this, the systems themselves need an accurate time scale to refer to - and most likely they rely on NIST.  So you'd be getting the NIST timebase, with noise.  On a human scale, over reasonable human periods of time, the errors are nil.  On a scale appropriate to today's computers, the story would be very different.  Deliberately manipulating the frequency across a whole grid for long enough to matter to humans would be extremely difficult and would get noticed:  We now have other, accurate time providers to compare our old electric clocks to.  Of course, if you're manipulating the environment, you can plug the device, not into the wall, but into your own frequency generator and make it see whatever you like.

The article Jonathan Thornburg linked to (http://www.bbc.co.uk/news/science-environment-20629671) describes the *forensic* mechanism.  It's based on looking at the short-time-scale variations in frequency around the nominal center point.  These are caused by variations in load, variations in supply (generators coming on and off line), lightning strikes, surges due to solar weather, and the interaction of these effects with the synchronizers that are put into the system exactly to keep those variations under control.  They are *not* local, but are constant within a single electric grid - synchronization across a grid is exactly the point!  Grids are very large.  England is covered by just one grid.  The continental US is covered by something like three, if I remember correctly.  (It may be a bit more, but we're still talking a handful.)  These variations are unpredictable in detail, but easily recorded anywhere on the grid.  Recording them *deliberately* as an absolute measure of time is an interesting idea - essentially aid the use the forensic technique.  In principle, one could probably replay past variations in a highly controlled setting to make it look as if a recording was made at some point in the past, but sounds rather hard to accomplish.  How successful one might be in replacing a recorded signal with a different one without leaving detectable artifacts is impossible to say without actual testing.

BTW, there's an interesting contrast here between a "tick generator" - which gives you an accurate repeatable way to step a clock from some fixed point - and an "absolute time reference", which lets you map from something (like a record of hum frequency variation) to absolute date and time.  We generally think of "clocks" as tick generators that we start off at some external date and time; absolute time references are relatively infrequent in day to day use.  (They are omnipresent in analyses of records of the past - e.g., looking at the stratum in which a fossil is found as a way of dating it.)  There's a Youtube video out there of a "clock" that measures elapsed time directly by looking at a couple of simple measurements that change as a potato rots....

                                                        -- Jerry

More information about the cryptography mailing list