Re: I've seen things you people wouldn't believe...

On May 8, 11:52=A0am, "David V. Loewe, Jr" <davelo...@[EMAIL PROTECTED]
>
wrote:
> On Thu, 8 May 2008 07:27:37 -0700 (PDT), petert...@[EMAIL PROTECTED]
 wrote:
> >On May 8, 12:00=A0am, "Keith F. Lynch" <k...@[EMAIL PROTECTED]
> wrote:
> >> David V. Loewe, Jr <davelo...@[EMAIL PROTECTED]
> wrote:
> >> > "Keith F. Lynch" <k...@[EMAIL PROTECTED]
> wrote:
> >> >> If you're using good quality encryption, they may still be able to
> >> >> tap it, but that doesn't mean they can tell what's being said.
> >> > If they REALLY want to know, they'll let the NSA play with it.
> >> > And they'll find out.
>
> >> I doubt that very much. =A0Computer power helps code-makers far more
> >> than code-breakers.
>
> >> > Trust me on this.
>
> >> Why? =A0Do you have inside knowledge? =A0If so, how come you're
allowed=

> >> to post about it?
>
> >Absent good reasons, I'm not going to trust David on this.
>
> >I have to agree with Keith here - while there is a heck of a lot of
> >lousy
> >security out there, I've never seen any evidence that the NSA (or
> >anyone
> >else) can break properly used modern crypto of adequate key length.
> >Breaches that do occur are due to poorly handled keys, poorly
> >implemented
> >software, or compromised systems. Doing it right is like maintaining
> >sterility in a medical operating field - a royal PITA, except you
> >won't know when you've failed until too late.
>
> >FWIW, I recently finished a 10-year stint as an engineer at
> >RSA Security Inc, implementing and deploying cryptgraphic software,
> >so this subject is familiar to me.
>
> 1) How much computing power does the NSA have at it's disposal?

Not enough - see below.

> 2) You do understand that Keith is talking about real time encryption
> of telephone conversations, correct?

Fine.

> 3) In light of 2), how many people have the capability of encrypting
> their phone *conversations*? =A0And who can they talk to?

Anyone with a PC or a Mac, and they can talk to anyone
with a PC or a MAC.

Check out, for example, Zfone: http://zfoneproject.com
a free, open source product. It's from Phil Zimmerman,
the guy who brought us PGP. It's an encrypted
VoIP system.

While this is using an ellipitc curve algorthim, Phil
states that the  licensable version has a strength
equal or exceeding that of 256 bit AES (and he's
a trusted expert in this area). Lets use that as
a starting point, but make everything else as
NSA friendly as possible.

We have 2^256 keys to test.

1. Let's suppose a 4GHz processor - fast, but
    not unbelievable.
2. Assume it can test 1 256 bit key every clock
   cycle. There are custom 'flow' key testing
   architectures which may be able to do this.
   You can't even get close on a general
   purpose processor.

   OK, with these optimistic assumptions, 1
processor could test all the keys in 2^222 seconds.

2. Lets get crazy, and assume the NSA is willing
to devote a 4 billion processor computer to the job.

3. No, lets be insane, and say they have 4 billion
computers, each with 4 billion processors, all
testing keys at 4 billion keys/sec (4 billion ~ 2^32,
which makes the maths easier).

The time to test all the keys comes down to a
mere 2^160 seconds.

That's over 4*10^40 years.

And that gets you the key for 1 (one) phone call.
You'll have to do it all over again for the next one.

Hmmm... checking - I forgot something! On
average you'll have to check only half the keys
before finding the right one - half the work!
So, revising down, it will take only about

20,000,000,000,000,000,000,000,000,000,000,000,000,000 years
per phone call, on average.

Sorry Dave, even Moore's Law won't help you here.

As John McCarthy said: "Those who refuse to
do arithmetic are doomed to speak nonsense."

BTW, I was the person behind the RSA
Symmetric Key Challenges, which succeeded in
cracking DES, and also cracked 64 key RC5.
I'm not speaking from a position of ignorance.

Peter Trei