Re: [Autoconf] I-D Action:draft-bernardos-autoconf-addressing-model-00.txt

[Ulrich Herberg <ulrich at herberg.name>]

Hi Teco,

On Wed, Oct 28, 2009 at 1:13 PM, Teco Boot <teco at inf-net.nl> wrote:

Hi Ulrich,

Few things.

Please realize your rich text mails is becoming a mess, when
replied with other tools. This troubles the already troubled
discussion. I stop cleaning up.

I did not realize that. If I just answer like now (without changing the font), is it readable? If not, I can also change to plain text for future mails.
 

Then on the small stuff: we are working on MANET stuff, OK?

I hope so :-)
 

We have a radio, right?

yes
 

We can use this radio for catching background noise, right?

yes
 

This is either thermal noise, human made noise, interference, timing,
timing of packets received, just to come up with a few.

Yes, I am aware that a radio can be good enough as entropy source. The question is: can we _always_ assume that it is good enough (or accessible to the device)? Is there an RFC that claims something like: "All devices that are used in a MANET MUST be able to provide a sufficiently good entropy source". (for whatever meaning of "sufficient")

Because, if we don't have that, I am not sure we can rely on good random numbers. Maybe we want to add this somewhere, or we have a common understanding that all devices have a good enough entropy source.

 

Please let me know this does not make sense. Because then, I want
to be warned and want to stay far, far away from such devices.

And still such devices might exist if we don't exclude their use in MANETs.
 

You were active on MANET and security. Spend some cycles on
PRN for small embedded devices with radios? I don't want to push you,
but it would help us if you do.

For security purposes, evidently the requirements for PRNs are very high. I admit that I am not an expert on PRNs, so I cannot tell you much. I will try to ask our cryptographers in the department, hopefully they can enlighten this issue.
To my knowledge, there are sufficiently good PRNs (even for small embedded devices), but they depend on very good entropy sources. So when specifying, say, a security extension to a MANET routing protocol, that document should probably mention some requirements of the entropy source and the PRN.


Ulrich

 

Regards, Teco





Van: Ulrich Herberg [mailto:ulrich at herberg.name]
Verzonden: woensdag 28 oktober 2009 12:43
Aan: Alexandru Petrescu

CC: Teco Boot; autoconf at ietf.org

Onderwerp: Re: [Autoconf] I-D

Action:draft-bernardos-autoconf-addressing-model-00.txt

Alex, Teco,
On Wed, Oct 28, 2009 at 11:54 AM, Alexandru Petrescu
<alexandru.petrescu at gmail.com> wrote:
Teco Boot a écrit :

[..]
The answer is: it depends. ;-) If you have a good random number generator
(i.e. using an unambiguous seed), then the collision probability is
extremely low. (I let you calculate it using the birthday theorem).

Are birthdays random ??


Probably not... I can imagine that there are some peeks 9 months after
extremely cold and rainy days ;-) I will not dig deeper into this issue :-)

 

However, if you have small embedded devices, without persistent memory, good
random number generator, etc., you can have collisions
with a much higher probability.  

Why? It is easy to build a pretty good PRN generator with radio HW.
I would say it is much, much, much, much easier than uniqueness
guarantee with try and error duplicate detection.

I agree.  Moreover, there is an RFC talking how to select good entropy
sources such that the PRN generator is good.

Do you mean RFC 4086? Yes, as you say, but all depends on selecting a good
entropy source. The RFC says:

"Is there any hope for true, strong, portable randomness in the future?
There might be. All that's needed is a physical source of unpredictable
numbers."

It also says that if no good hardware entropy sources are available, "there
are other possibilities. These include system clocks, system or input/output
buffers, user/system/hardware/network serial numbers or addresses and
timing, and user input. Unfortunately, each of these sources can produce
very limited or predictable values under some circumstances."

So it really depends on the kind of hardware we are talking about. I could
imagine very small sensors without persistent state or other hardware source
such as mentioned in the RFC. Using such a device could lead to collisions
when calculating random numbers.


(small embedded devices, such as sensors, can constitute good sources of
 randomness when sampling the temperature, for example - that RFC says).

yes. But can we guarantee that each MANET router has a good entropy source?
Please point me to some reference saying that each device that can be used
in a MANET must have a good entropy source.

Ulrich



Alex
Not even
mentioning total network meltdowns after a first MANET merge.

And I don't have those devices, I want to stay far, far from them.
And I don't want to run protocols for it in my MANETs.
Please provide refs to equipment that you have in mind. First,
I'll check why good PRN can't be supported. If not, I can stay far away from
these (also having security in mind).

Still waiting on answers on DHCP DUID, key generation etc.


Regards, Teco
_______________________________________________
Autoconf mailing list
Autoconf at ietf.org
https://www.ietf.org/mailman/listinfo/autoconf