Re: [dhcwg] Order of options for DHCP Anonymity profile

Ordering by option number seems a fine (and simpler) approach. The differences in options requested might still make fingerprinting possible (especially for clients that ask for more unusual options).

Perhaps for anonymity, it might even be appropriate to consider reducing the option set a client asks for to those it really needs (for basic connectivity). Though this might also have minimal impact, as I suspect the more bizarre option requests are typically from things that aren't usually very mobile?

-          Bernie

From: dhcwg [mailto:dhcwg-bounces@ietf.org] On Behalf Of Sten Carlsen
Sent: Sunday, August 30, 2015 2:11 PM
To: Christian Huitema; dhcwg@ietf.org
Subject: Re: [dhcwg] Order of options for DHCP Anonymity profile

On 30/08/15 19:08, Christian Huitema wrote:

In his review of the anonymity profile, Bernie suggested considering the order of options and the PRL parameter of DHCPv4:

BV> New section between 3.4 and 3.5 to include the PRL similar to ORO text in

3.5? Kind of odd that the Parameter Request List isn't mentioned at all for

DHCPv4 where there exists a fairly extensive fingerprinting database (mostly

based on the PRL).

I am looking at the current text in the draft, and the DHCPv6 paragraph reads:

   One specific method used for fingerprinting utilizes the order in

   which options are included in the message.  Another related technique

   utilizes the order in which option codes are included in an Option

   Request Option (ORO).

   The client willing to protect its privacy SHOULD randomize options

   order before sending any DHCPv6 message.  Such a client SHOULD also

   randomly shuffle the option codes order in ORO.

We could add similar text in the DHCPv4 section, but I have a small question regarding practicality of "randomizing the options." The current code is simple and static, with different blocks processing different related options. Randomization would require either executing the code blocks in random order, which is tricky, or adding a post processing phase to shuffle components of the messages, which is not natural.

As far as the ORO or PRL options are concerned, I would suggest that requested option numbers be sorted from smaller to larger. That allows for static implementations, and defeats fingerprinting just as well as randomization.

For the ordering of the actual options, I don't think that requiring randomized order is practical, and I would prefer allowing implementers to use a static order.

Any advice from the WG?
This is a tricky issue. There are several paths, each with its perils:
- some devices on the network randomises, some don't - this highlights who wants to be anonymous.
- the randomisation function might be recognisable, yielding another fingerprint.
- the collection of what you ask for is still a good part of a fingerprint regardless of sequence.

The most anonymous approach would probably be that each device asks for the exact same things in the exact same order all over the world. This is obviously not going to happen.

Just my 0.02$

-- Christian Huitema

_______________________________________________

dhcwg mailing list

dhcwg@ietf.org<mailto:dhcwg@ietf.org>

https://www.ietf.org/mailman/listinfo/dhcwg

--

Best regards

Sten Carlsen

No improvements come from shouting:

       "MALE BOVINE MANURE!!!"