[gnso-rds-pdp-wg] The principle for thin data (was Re: Principle on Proportionality for "Thin Data"access)

Thu Jun 1 16:26:57 UTC 2017


Sent from my iPhone

> On Jun 1, 2017, at 09:43, Stephanie Perrin <stephanie.perrin at mail.utoronto.ca> wrote:
> 
> I certainly agree that if people enter personal information as part of their DNS registration or their motor vehicle licence registration, it is done with implied consent... as long as there is sufficient information to permit them to understand just how the data is being used and where it is going.  However, as I tried to say with respect to registering a domain name, I really don't think the average non-expert citizen who might want to register a domain name would get enough information to truly understand how far his/her information goes, and how difficult it is to get it removed once it has appeared in the public record.  We should build this system so that everyone understands it, not just the experts.
> 
> cheers Stephanie
> 
>> On 2017-06-01 05:18, jonathan matkowsky wrote:
>> Stephanie,
>> 
>> I agree with you that we should not conflate collection limitation principles with openness principles.
>> 
>> I respectfully disagree with most of what you wrote in the first paragraph of your post script.   Here we are talking about users potentially entering personal or pseudonymous information when they are not being asked for it (nor is it required) to begin with, and it is not required for purposes of which it's being collected. That is the   scope  of what needs to be assessed  if at all and how the scope needs to be  defined from the beginning  if you were to conduct a PIA .  
>> 
>>    Personal information is not being used or intended to be used just because a person decides to enter personal information into a field.                        The example of how you can combine databases to re-identify a person based on the SOA record is the equivalent of protecting domain names as personal information because a person  can register their driver's license  or name and date of birth as a domain name.   I would argue no PIA should be required  as a result  even in accordance even with best practices.   A PIA needs to be conducted in a manner that is commensurate with the level of privacy risk identified . 
>>  
>> I respectfully disagree with you that thin data is personal. We are talking about identifiers (codes or strings that represent an individual or device).  Many labels can be used to point to individuals. Some are precise and most, imprecise or vague. There's no question that an IP address is a device identifier.  Device IDs, MAC addresses can be a source for user tracking.  But  i dentifiers can be strong or weak depending on how precise they are as well as the context. It cannot be measured without taking linkability into consideration.  For that reason, name servers are not the same as IP addresses or MAC addresses any more so than the existence of a domain name is an identifier. If a person chooses to use identifiable information when it is not being asked for or required for purposes of which the data is being collected, that does that mean we need to classify all the data according to that unlikely scenario. Those setting up their own DNS would be relatively speaking, sophisticated Internet users that presumably know the basics of how DNS operates in any case, so by entering the information in that way, they are choosing to customize their DNS in a personal way similar to a person that chooses to show personal information on their license plate number.  
>> 
>> I know that the motor vehicle registry is restricted now in most places so that you would need a subpoena to get that kind of personal information. This is also true of an IP address though and IP providers. The fact is a person can put their name and date of birth on a license plate if they want to customize it. And then they get on the road. That does not mean the license plate numbers are all personal information. It's pseudonymous data. It is true that it is a stronger identifier than an IP address insofar as if you subpoena the motor vehicle registry operator, you will get the personal information behind that license plate number. If you subpoena the ISP, you MIGHT get the personal information depending on the nature of the IP address. It's still true that to drive a car, you need to show your license plate number on the vehicle. 
>> 
>> I would argue that thin Whois data is pseudonymous or personal data to the same extent that a person can choose to customize a license plate if they want to, and               put personal or psuedonymous data into fields for which the data being collected does not ask for or require them to do so.   
>> 
>> A  person can register their driver's license as a domain               name. They can use a personal email in their SOA record, or personal NS.   Just because it's theoretically possible for someone to enter pseudonymous (or even personal) data into multiple databases when they are not being asked for it, and those combination of choices make it possible to identify them, does not mean one of the sets (Thin Whois) should be classified as personal information subject to a PIA. 
>> 
>>  
>> Jonathan Matkowsky,
>> VP – IP & Brand Security
>> USA:: 1.347.467.1193 | Office:: +972-(0)8-926-2766
>> Emergency mobile:: +972-(0)54-924-0831
>> Company Reg. No. 514805332  
>> 11/1 Nachal Chever, Modiin Israel
>> Website
>> RiskIQ Technologies Ltd. (wholly-owned by RiskIQ, Inc.)
>> 
>> On Thu, Jun 1, 2017 at 12:02 AM, Stephanie Perrin <stephanie.perrin at mail.utoronto.ca> wrote:
>>> Your summary today was great Andrew.
>>> 
>>> I am not arguing about the disclosure of thin data.  We already voted on unauthenticated mandatory disclosure, weeks ago (or at least it feels like weeks ago).  Lets please move on.  We are debating this yet again, because people keep asking, is thin data personal?  [lots of people missed the last call]  The answer is yes (IMHO).  Does that mean it cannot be disclosed?  The answer is no.  Does the proportionality principle apply?  Yes.  Have we already gone through this?  Yes.  Can we come back to it?  Yes, but hopefully only if we have to.....we will have to when we get to data elements.
>>> cheers Stephanie
>>> PS a fundamental problem here is that people try to categorize information that in their view should be disclosed, as not personal information.  This fight has gone on for years over IP address, for instance.  The important question is not actually whether it is personal data or not, it is "do you need to disclose it to make things work?"....and if the answer is yes then you try to mitigate the disclosure and try to keep it minimized to what is absolutely required.  Hence the PIA, which should employ both data minimization and the test in the proportionality principle as techniques to evaluate data elements.
>>> A good and really simple example is a phone number.  IS it personal info?  (the telcos fought for years, trying to claim they owned it and it was not personal).  Obviously it pertains to you, people feel strongly that it is personal (culturally relative of course but...) and yet if noone ever learns your number your phone won't ever receive a call.  That does not mean you have to disclose it everywhere.....only where necessary.  And it should mean that it does not have to follow you everywhere, but that is becoming increasingly hard to manage....
>>> 
>>> By the way, informed consent is not the same as transparency requirements.  Transparency requirements are exactly that....you have to be transparent about what you are doing with data.  Let us not conflate that with consent.
>>> 
>>> I will quit now and stop trying to answer questions.  I would like to humbly suggest, however, that we have a real shortage of basic understanding of how data protection law works and is interpreted.  If there is a data protection law expert that folks might listen to, we should hire that person to advise us.  It might save a lot of time.
>>> 
>>> 
>>>> On 2017-05-31 16:00, Andrew Sullivan wrote:
>>>> Hi,
>>>> 
>>>> On Wed, May 31, 2017 at 03:20:59PM -0400, Stephanie Perrin wrote:
>>>>> That does not mean we need to protect it, it means we have to examine it in
>>>>> terms of DP law.  May I repeat the suggestion that Canatacci made in
>>>>> Copenhagen in response to a question.....(I forget the precise question he
>>>>> was asked, sorry). If you want to figure out whether you have to protect
>>>>> something or not, do a privacy impact assessment.
>>>> As I think I've said more than once in this thread, I think we _have_
>>>> done that assessment and I think the answers are obvious and I think
>>>> therefore that there is nothing more to say about this principle in
>>>> respect of thin data:
>>>> 
>>>>     - the data is either necessary for the operation of the system
>>>>       itself or else necessary for distributed operation and
>>>>       troubleshooting on the Internet.
>>>> 
>>>>     - the data does not expose identifying information about anyone,
>>>>       except in rather strained examples where the identifying
>>>>       information is already completely available via other means.
>>>> 
>>>> What more is one supposed to do? 
>>>> 
>>>> Best regards,
>>>> 
>>>> A
>>>> 
>>> 
>>> 
>>> _______________________________________________
>>> gnso-rds-pdp-wg mailing list
>>> gnso-rds-pdp-wg at icann.org
>>> https://mm.icann.org/mailman/listinfo/gnso-rds-pdp-wg
>> 
> 
> _______________________________________________
> gnso-rds-pdp-wg mailing list
> gnso-rds-pdp-wg at icann.org
> https://mm.icann.org/mailman/listinfo/gnso-rds-pdp-wg
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mm.icann.org/pipermail/gnso-rds-pdp-wg/attachments/20170601/06a2443c/attachment-0001.html>