That works for me… it's compatible with further discussion that doesn't result 
in changes to the draft's scope of contents.

R

Robin Wilton
Technical Outreach Director - Identity and Privacy
Internet Society

email: [email protected]
Phone: +44 705 005 2931
Twitter: @futureidentity




On 10 Aug 2012, at 10:43, Rhys Smith wrote:

> My 2c - as I said in my last message, I think the most the draft should do is 
> mention that data perturbation is an option, and give links to further 
> material. When we discuss anonymisation we don't go into detail about the 
> various methods that can be used to achieve it. We shouldn't with this 
> either. The draft is not about defining *how* to achieve privacy, but to 
> discuss what privacy is, give an overview of ways of protecting it (but not 
> in any detail), and to provide guidance on how to talk about privacy.
> 
> There are various methods available to achieve the effect of "fuzzing" data, 
> each of them fairly dependent on the domain in use. E.g. algorithms to 
> achieve it across large data sets to achieve privacy of single tuples are 
> completely different to those that achieve geographic privacy for a user for 
> geo-located data which are completely different to those attempting to 
> achieve privacy by putting false traffic into something like an anonymisation 
> network (e.g. Tor) to hinder traffic analysis, etc etc etc. So a full 
> discussion about it would probably be a full journal paper in its own right. 
> In fact, a full discussion about data perturbation per domain is a journal 
> paper in its own right, or indeed, a whole class - cf. 
> http://theory.stanford.edu/~nmishra/cs369-2004.html for statistical 
> databases, the geopriv discussions for geographic information, etc.
> 
> I'm very wary of introducing major new topics into the draft, given past 
> feedback has been that it's already too complex. The privacy considerations 
> draft needs to kept as short and snappy as possible if it's going to have any 
> chance of the IETF community reading it, understanding it, and using it.
> 
> So my opinion, just to restate - is that the draft should mention that data 
> perturbation is an option, and give links to further material. Couple of 
> paras at most.
> 
> R.
> --
> Dr Rhys Smith
> Identity, Access, and Middleware Specialist
> Cardiff University & Janet - the UK's education and research network
> 
> email: [email protected] / [email protected]
> GPG: 0xDE2F024C
> 
> 
> 
> 
> 
> On 10 Aug 2012, at 09:58, Robin Wilton wrote:
> 
>> Hmm - but unacceptable to whom? There are definitely times when I am 
>> perfectly comfortable self-asserting a false location. In fact, I'd go 
>> further and say that in general, I have no use for location-based services…
>> To be honest, I think service providers often get the location-based 
>> services argument the wrong way round; what's useful to me, as a user, is 
>> the ability to go online and locate something (say, a restaurant) regardless 
>> of *my* current location. (So, for instance, I can find out where my hotel 
>> in San Diego for next week is, even though I'm in the UK). I am less 
>> interested in passively disclosing my location so that I can be told what is 
>> in my immediate vicinity.
>> 
>> I take Martin's point about location fuzzing: the fact that I state a false 
>> location on Twitter won't fool someone who carefully monitors the times at 
>> which I tweet… they will quickly figure out that either I often tweet at 3 
>> in the morning, or I'm not where I claim to be. But I think we should be 
>> careful about how we frame the problem and the potential goals. It is almost 
>> certainly not realistic to aim make it impossible for anyone to de-identify 
>> any data about me: in privacy terms, it's more viable to aim to raise the 
>> threshold of data needed for *some* third parties to infringe my privacy. 
>> The EU Article 29 Working Group implies this with its findings on what 
>> constitutes personal data. Their view is that some items of data (such as an 
>> IP address) are sometimes personally identifiable and sometimes not, 
>> depending on whether a third party is in a position to link them to other 
>> data items. Thus, as far as my ISP is concerned, my IP address is easy to 
>> link with a subscriber address… whereas to most other third parties, it is a 
>> relatively fuzzy identifier.
>> 
>> Bottom line: I'm not sure if the required action here is further research 
>> work or changes to the draft, but I do think the problem would benefit from 
>> being explored and defined more fully…
>> 
>> HTH,
>> Robin
>> 
>> 
>> Robin Wilton
>> Technical Outreach Director - Identity and Privacy
>> Internet Society
>> 
>> email: [email protected]
>> Phone: +44 705 005 2931
>> Twitter: @futureidentity
>> 
>> _______________________________________________
>> ietf-privacy mailing list
>> [email protected]
>> https://www.ietf.org/mailman/listinfo/ietf-privacy
> 

_______________________________________________
ietf-privacy mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/ietf-privacy

Reply via email to