Re: [Cfrg] Curve selection revisited

[Robert Moskowitz <[e-mail]>]

The whole world is NOT TLS.  For example I am working in IEEE 1609 which 
is Vehicle-to-vehicle safety communications and their need is for signed 
data objects that are quickly generated and verified (you have ~100ms 
from the braking event to all the cars being able to act on that 
information).  I also work in IEEE 802.15.4 highly constrained devices 
(some much smaller than what many IETF groups accept as their lower limit).

8 bit processors are the norm, and NO, Dr. Moore is NOT going to change 
that.  Lower energy demand is much higher on the design plan than faster 
processing.  Also whatever is done will be running on fielded devices 
for the next 10 - 20 years.  Hardware acceleration is acceptable, but 
that could make in-field corrections impossible. Over-the-air firmware 
updates may also be impossible (802.15.4k is 40kb/s) or only done in the 
dealership (yes!  another recall!).

So getting on to my thoughts of curve selection requirements for the 
world of small:

1) Low over-the-air demands.  Every BIT counts.  We have to live with 32 
bytes; but 64 bytes may be too much.  Oh, IEEE 1609.2 has its own 
'certificate' and signature format as even the standard asn.1 is too 
wasteful of the very limited channel.  This applies also in 802.15.4 
where adding another MAC frame might mean that the whole thing fails 
because now it is more likely that one frame got lost.

2) Easy/size of implementation.  There has been lots of chatter here 
about how NIST p256 can be implemented wrongly.  Here it is of greater 
concern, as it might require a physical swap-out to correct an 
implementation error.  Manufacturers look at the cost of a programmer to 
save 5K code cheaper than that extra 5K footprint.  So little code and 
easy to code.  This world is extremely price sensitive.

3) Performance on an 8 bit  processor.  Performance is measured in time 
and energy.  In-vehicle devices tend to have a ready source of electrons 
(except for tire gauges!), but some devices have limited battery (your 
front door lock) or NO battery (MEMEs energy harvesters on bridge stress 
sensors).  So low energy is as important as quickly done.  Here 10% 
faster IS 10% faster; and it could make a life-or-death difference. (In 
Positive Train Control, the automatics MUST take over if the human can 
no longer respond in time and bullet trains can take 7Km to stop safely)

4) preserving ECDSA and ECDHE.  This is what is currently in use, and in 
some cases written into regulations (urgh), so we have to still keep 
them working.  But if something 'new' and provably good makes a big 
difference in one of the previous needs I suspect there will be an 
attentive audience.

I have tried to keep things general so as not too much to limit 
choices.  I never claimed to be a crytographer or to have the math 
skills to evaluate the various contenders.  Others working in this area 
might have their own reading on the needs of constrained 
devices/communications.

But please just don't solve this for TLS on octo-core 64-bit systems.


On 07/25/2014 01:10 AM, Benjamin Black wrote:
> Inspired by some comments from other and in hopes of returning to a 
> more productive debate rather than the pointed back and forth, of 
> which I am absolutely guilty, I'm suggesting a few constraints. These 
> are not priority ordered and none of them deal with the math, instead 
> focusing on process and practice.
>
> 1) As we are discussing this in the context of TLS, the performance of 
> various kx options must be evaluated for ECDHE as typically deployed. 
> ECDH is neither common nor recommended in the TLS BCP draft. In 
> addition, I hope there will be discussion of benchmarking methodology, 
> with the understanding the ultimate choice is for the CFRG chairs.
>
> 2) Performance must be measured using "production-quality" 
> implementations. By this I mean implementations which employ the sort 
> of techniques/optimizations appropriate for large scale deployment. 
> This is specifically intended to exclude discussion of how simple or 
> fast an implementation _could_ be, in favor of what they actually are 
> in practice. However, the goal is to select curves which strike the 
> best balance between various requirements, not simply the fastest.
>
> 3) The timeframe for a decision constrains the scope of this process. 
> If a decision is desired within a few months, then it is difficult to 
> include options beyond new curves. If a decision can take a year, new 
> algorithms could be considered. Given the importance, impact, and 
> contentiousness of new algorithms, I believe it would be difficult to 
> complete a thorough, careful algorithm and curve selection process in 
> a short amount of time.
>
> 4) The precise set of security levels needed should be specified in 
> advance (whether by TLS or CFRG) and curve recommendations delivered 
> together for all levels. A stated goal is to use these curves in new 
> drafts, and it minimizes work to specify all security levels for a 
> proposal in the same draft. This is strictly a practical matter 
> recognizing the chairs and area directors have limited bandwidth and 
> implementors can get everything done in a single go.
>
> 5) Selections must efficiently support TLS 1.2. Adoption of new 
> curves, and potentially new algorithms, can't be gated on completion 
> and widespread deployment of TLS 1.3.
>
> 6) Drop-in replacement for the NIST curves is _not_ a requirement in 
> this process.
>
> 7) I don't think the chairs are signing up for a full-on, NIST-style 
> selection competition and we should aim to keep things are lightweight 
> as possible while still achieving the required level of rigor. I hope 
> they will address this point soon.