Re: [ANCP] Privacy issue in draft-ietf-ancp-mc-extensions-12

[Tom Taylor <[e-mail]>]

On 02/12/2013 10:15 AM, Ted Lemon wrote:
> I think that I've failed to adequately communicate the concern I have
> about this.   It might help for folks to review
> http://tools.ietf.org/html/draft-ietf-6man-ipv6-address-generation-privacy-00,
> particularly section 3.1 (but you should probably read up to that so
> that you get the context).
>
> I am assuming here that the device whose MAC address is being
> identified is some end-user device, not the home gateway.   On an
> IPv4 network, or an IPv6 network for devices that use stable privacy
> addresses
> (http://tools.ietf.org/html/draft-ietf-6man-stable-privacy-addresses-15),
> there is no ability to correlate activities based on the MAC address
> of the host.   So using the MAC address as an identifier for access
> control goes against the current trend in the IETF of avoiding using
> global, fixed identifiers like the MAC address in this way.
>
> If there is a need for an identifier to be used for access control,
> it should be an identifier specific to the context, which will not be
> used in other contexts.   So it shouldn't be the MAC address.
>
>
[PTT] Point about MAC address privacy accepted. Thinking over the IP 
address alternative, I can see practical problems -- since the device is 
in the home network, the provider has no control over the stability of 
the address (if IPv6), and can't determine the device from the address 
(if IPv4 with the usual NAT at the home gateway). My proposed technical 
solution would be to require the AN to create a context-specific device 
identifier by hashing the MAC address with something stable at the AN, 
with the result predictable so that the subscriber AAA profile can be 
set up as required.

The use cases I could see for this are not essential but possible:
-- provider-enforced parental controls on selected devices
-- provider-enforced device-specific limitations (e.g., mobile vs. fixed)

Have to admit the second case at least seems very weak.

Tom Taylor