Re: 64bit MAC addresses and SLAAC

[Alexandre Petrescu <[e-mail]>]

Le 17/06/2020 à 23:54, Philip Homburg a écrit :
>> Of course the caveat there with unmanaged network and SOHO and
>> Mobile where manual or DHCPV6 is not possible or viable.
> 
> Why is DHCPv6 not possible? Of course there is the Android problem. 
> But I doubt that a low power IoT device would share a wifi-link with
> an Android device.
> 
>> In those cases SLAAC is preferred, but then we have the crux of 
>> issue and the decision tree on privacy random IID and its overhead 
>> if its not necessary versus modified EUI64. tree of course the
>> underlying operational impacts of random versus  stable IID double
>> edged sword operator or individuals decision to pick which works
>> best for their use case.  In the end net-net is what is simplest to
>> deploy and least overhead but also meets the desired goal is
>> generally the thought for picking the IID generation solution. For
>> that SLAAC wins out in that decision for the use case described
>> above.
> 
> A stable, unique pseudo random IID per prefix basically requires one 
> SHA2-HMAC (or equivalent in SHA3 or other modern hash function) per 
> prefix.

I am not sure which crypto function is used in our (well, a colleague
programmer's) linux kernel code to generate variable length IIDs.  We
generate a 128bit random number with a crypto function and then pick the
necessary bits to form an IID from it and from the non-64 prefix from
PIO from RA.

However, by doing that we become fully dependent on that crypto function
(say it's called 'C') to generate an address.

If that 'C' is not seeded well enough, or has licenses on it, or is
IPRed, or is not open source, or bleeds when attacked by a
distributed.net lucky PC, or is not resisting quantum attacks after
tomorrow, or was not fully tested against endianness on this new IoT
platform, or is too slow, or simply does not compile there - then we're
at risk.

Alex

> 
> I cannot imagine that computing 2 SHA2 hashes has a significant
> amount of energy use compared to link attachment and duplicate
> address detection.
> 
> This computation is once per prefix, so it becomes an issue if the 
> devices frequently attaches to new links. In which case tracking
> protection becomes important.
> 
> I also wonder, a device that cannot afford a couple SHa2 hashes? Then
> that device is probably insecure and should not be connect to the
> internet anyhow.
> 
> 
> -------------------------------------------------------------------- 
> IETF IPv6 working group mailing list [e-mail] Administrative
> Requests: https://www.ietf.org/mailman/listinfo/ipv6 
> --------------------------------------------------------------------
>