Thanks all for a thorough back and forth on this! I find the conversation has moved on in my day's absence ! Reviewing the thread, I do find myself in line with the NCSG viewpoint of this, and indeed with Alan's last statement on this. 1) Identification of the Data Subject is no longer limited to from within the data set of the controller, but whether there are other avenues, which in the time of 'big data' and 'advanced research methods', can identify the data subject from other data also available out there. The risk can be certainly lowered, where that other data is not ordinarily available, or the data is beyond the reach of such research methods, but when we are talking about registrants of domain names, and where such domain names usually are associated with websites, there is a much larger likelihood that identification or the data necessary to identify the data subject is present. [Need I remind that in this sphere, all it takes is 1 complaint from an aggrieved data subject to start the cycle of pain] For our purposes that's enough, there are currently too many variables and potentials, and whereas 'big hitter' like Microsoft may (perhaps erroneously) deem their own risk is acceptable, that's their choice, but we cannot take the choice away. Smaller more financially exposed CPs might not be so 'sure'. The simple fact is where we are unable to demonstrate how we actually control the risk, we cannot make policy tending towards the reckless enhancement the risk of the CPs. 2) Ayden has provide a very solid example of how the law has evolved on this matter, and I must disagree with Hadia's interpretation of both the GDPR and ECJ jurisprudence on this. I believe Ayden's interpretation to be more closely aligned with the Court's stance. 3) Alan G's final point is that "*Why not leave it to the registrar to decide*". I believe, as we are prone to do in such conversations, come full circle. I therefore restate my suggested permissive wording, allowing for a registrar (or registry) to publish, where like for instance Microsoft, that they believe that their risk is acceptable. RECOMMENDATION XX *In the case of a domain name registration where a privacy/proxy service used (e.g. where data associated with a natural person is masked), Registrar (and Registry where applicable) MAY include in the public WHOIS and return in response to any query full WHOIS data, which may also include the existing privacy/proxy pseudonymized email.* (NOTE: to your original point Alex, this does not render the recommendation useless,as our experience with ICANN compliance is such that what ICANN policy permits, is equally as important as what policy prohibits .) Great conversation all! Alan [image: Donuts Inc.] <http://donuts.domains> Alan Woods Senior Compliance & Policy Manager, Donuts Inc. ------------------------------ The Victorians, 15-18 Earlsfort Terrace Dublin 2, County Dublin Ireland <https://www.facebook.com/donutstlds> <https://twitter.com/DonutsInc> <https://www.linkedin.com/company/donuts-inc> Please NOTE: This electronic message, including any attachments, may include privileged, confidential and/or inside information owned by Donuts Inc. . Any distribution or use of this communication by anyone other than the intended recipient(s) is strictly prohibited and may be unlawful. If you are not the intended recipient, please notify the sender by replying to this message and then delete it from your system. Thank you. On Tue, Feb 5, 2019 at 2:04 AM Alan Greenberg <alan.greenberg@mcgill.ca> wrote:
Ayden, Why not leave it to the registrar to decide. We have heard an awful lot about potential for contracted party risk. They can make their own decision. And those who have already implemented secure anonymized e-mail do not have to scrap it and implement something completely new, as they would if we follow your suggestion.
Alan
At 04/02/2019 08:00 PM, Ayden Férdeline wrote:
Hi Alan G.,
I happen to agree that this could be an implementation issue. However, we must balance the cost of implementing such a complex system and rolling it out globally against the benefits to registrants, who are ultimately those who will bare its cost. In light of this consideration, and in recognizing the harm that could arise from an improper implementation, I cannot find myself supporting this proposal. I see no benefit to the registrant of the publication of a pseudonymised email that would not otherwise be achieved by making them contactable through a web form.
Kind regards,
Ayden Férdeline
†††††††Original Message †††††††On Monday, February 4, 2019 7:19 PM, Alan Greenberg < alan.greenberg@mcgill.ca> wrote:
That *may* apply in some cases but certainly not all. There are European registrars that use anonymized addresses that change regularly (such as weekly, daily or oftener). There is no way to correlate them with previously published data.
It also addresses the worry that the anonymized address can be used for SPAM.
This is an implementation issue. Each registrar can use a technique which meets their needs and reduces their risk.
Alan
At 04/02/2019 06:20 PM, Ayden Férdeline wrote:
I agree with Milton̢۪s interpretation.
We also have case law that can help us understand the meaning of where pseudonymisation is “reasonably likely†to lead to the ide identification of a data subject.
In *Breyer v Germany*, the Court of Justice of the European Union̢۪s ruling mirrors much of the de debate we are having here, and it found that pseudonymized data is almost always personal data. (While this case was a judgment on the 1995 Data Protection Directive and not the GDPR, the text of Recital 26 in the GDPR and in the Directive differ very little in substance.)
In light of this, and given the relative ease with which historical WHOIS records can be retrieved, I believe it is only prudent that pseudonymised email addresses are not published.
Best wishes,
Ayden
††††††¬ †Original Message ††â€Ã¢€ †††On Monday, February 4, 2019 1:57 PM, Mueller, Milton L <milton@gatech.edu> wrote:
Marc:
I have paraphrased this as:
“If a processor were to publish or disclose multltiple personal data, one of which was a pseudonym, and if that collection of data could be used to render the pseudonym identifiable, then the pseudonym would be considered to be personal data. However, if the processor were to publish a pseudonym, and a third party were to subsequently locate additional data, available from different processors or even from the same processor in a different context, and were then able to use these separate data to unmask the pseudonym and render it identifiable, such third party correlation ability would NOT require the processor to treat the pseudonym as personal data.â€
I don̢۪t find your lawyers̢۪ opinion very conviy convincing, Marc. She is saying that even if it̢۪s knownown for a fact that the pseudonym can be combined with other data elements that are readily accessible to anyone to identify the subject, it is not personally identifiable simply because all the data isn̢۪t in one place. Th That doesn̢۪t pass the giggle test.
* Second, the Whois DOES publish additional, multiple personal data, such as country, state, city, nameserver, domain name, etc. all of which when combined could be used to identify. Dr. Milton L Mueller Professor, School of Public Policy <http://spp.gatech.edu/> Georgia Institute of Technology Internet Governance Project http://internetgovernance.org/ <http://internetgovernance.org/> /marksv From:* Gnso-epdp-team < gnso-epdp-team-bounces@icann.org> *On Behalf Of *Alex Deacon *Sent:* Saturday, February 2, 2019 11:57 AM *To:* Alan Woods <alan@donuts.email> *Cc:* EPDP <gnso-epdp-team@icann.org > *Subject:* Re: [Gnso-epdp-team] Recommendation on privacy/proxy data.
Thanks Alan. As always I appreciate your input and views. A few thoughts and questions on your proposed solution.
First, given all of the "MAYs" in your suggested update, it is not clear to me how this will be translated in the implementation phase. Specifically it is a no-op from a compliance point of view. I understand this may be the main reason for your update, but we end up with a Recommendation with little value.
Second, given the general agreement that we avoid "squishy" language in our purposes and recommendations, the language you propose is unclear (to me at least). If a Registrar or Registry decides to not return P/P data or the "existing p/p pseudonymized email†does that mean no email adddress will be available in a public response? If so that seems to go against our existing Contractibility purpose (Purpose 3). (Alan G made this point I believe.) Or does this language assume that there will be a Registrar pseudonymized email that points to a P/P pseudonymized email? Both seem problematic.
Third, regarding the language you point out in Recital 26 of the GDPR, its not clear to me it applies in the privacy/proxy context, where the only the psuedonymized email could be seen as personal. Perhaps this would be a good question to pose to Ruth. From what I understand if a processor were to publish or disclose multiple personal data, one of which was a pseudonym, and if that collection of data could be used to render the pseudonym identifiable, then the pseudonym would be considered to be personal data. For responses that only included P/P data this wouldn't be the case.
Fourth, I appreciate your concern that it may not be possible to figure out a P/P provider and suggest we can use the language in the RAA (Section 2) to be more specific here - essentially limit it (for now) to affiliated services. e.g. “For any Proxy S Service or Privacy Service offered by the Registrar or its Affiliates, including any of Registrar's or its Affiliates' P/P services distributed through Resellers, and used in connection with Registered Names Sponsored by the Registrar,". In the future, and once the PPIRT completes, all accredited P/P services will be flagged in RDS/WHOIS (PPSAI Recommendation 4) so this issue will be solved.
Finally, I continue to believe we can find a pragmatic solution where we eliminate (or perhaps minimize) the situation where P/P data is returned in response to a "reasonable disclosure" request. It is a very inefficient use of time/cycles for both the requestor and person responsible for manually processing these requests. This was the main reason for the Temp Spec language and this new Recommendation. If we can't address this issue in this new Rec then perhaps language needs to be added to Rec 12 to address this case....
Apologies for the long email.
Alex
___________
*Alex Deacon*
Cole Valley Consulting
alex@colevalleyconsulting.com
+1.415.488.6009
On Fri, Feb 1, 2019 at 4:47 AM Alan Woods <alan@donuts.email> wrote:
Thank you Alex for reopening this matter.
From a practicality POV the inclusion of a pseudonymised email in a publication is a disclosure of public data, as a pseudonymised email, is still an email for a data subject. It may be in a form the data subject does not recognize, but I believe a number of excellent and practical examples, such as auto-relay / auto out of office responses could easily identify the non-pseudonymised email and lead to identification of the Data Subject.
I refer you recital 26 of the GDPR:
"…Perrsonal data whiich have undergone pseudonymisation, which could be attributed to a natural person by the use of additional information should be considered to be information on an identifiable natural person..."
Alas the recommendation as worded, although I understand the practicality that is driving it, does not properly address data protection concerns. There is a lot of back and of forth possible on whether pseudonymisation may count as anonymization when published to a 3rd party, however in the context of Domain Names, where the domain name, and even the associated content, could be capable of linking the individual to the pseudonymised data (which is completely outside of outside of our control, but still relevant to our risk assessment), we need to tread carefully; more carefully than the ePDP has either the time or the scope to consider.
To refer again to our goal : We must look to the state of the data that is in the RDDS currently, and whether the publication of individual data currently used would be data protection compliant or not. In its current state, with the myriad the uncertainties, it is not, therefore our recommendation must be not to publish.
SOLUTION
Allow the registrar / registry to consider (as Controller) the feasibility and the risk as it applies to them? Unless we, at this late stage, the ePDP can provide an actual means or suggestion as to implementation of how a registrar / registry can effectively figure out who is a P&P provider and thus publish those only details without an increase to risk, then our choice is plain.
If the recommendation is to stand, use of MAY, and other permissive language, as opposed to a 'must' will allow each CP to assess the risk as they see it, as it all comes down to the fact that the ePDP cannot force the CPs to assume higher risk to appease some.
Recommendation XX
In the case of a domain name registration where a privacy/proxy service used (e.g. where data associated with a natural person is masked), Registrar (and Registry where applicable) MAY include in the public WHOIS and return in response to any query full WHOIS data, which may also include the existing privacy/proxy pseudonymized email.
Kind regards,
Alan
Senior Compliance & Policy Manager, Donuts Inc. <https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fdonuts.domains&data=02%7C01%7Cmarksv%40microsoft.com%7Ccd7ec2e952744fc76eb608d68948bdc7%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636847342852603079&sdata=QwaxAMupO7ehwbSmN9I%2FFKwgNNr%2FP0th2kuQW6EHb9U%3D&reserved=0>
------------------------------
15-18 Earlsfort Terrace <https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fdonuts.domains&data=02%7C01%7Cmarksv%40microsoft.com%7Ccd7ec2e952744fc76eb608d68948bdc7%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636847342852603079&sdata=QwaxAMupO7ehwbSmN9I%2FFKwgNNr%2FP0th2kuQW6EHb9U%3D&reserved=0> Dublin 2, County Dublin <https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fdonuts.domains&data=02%7C01%7Cmarksv%40microsoft.com%7Ccd7ec2e952744fc76eb608d68948bdc7%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636847342852603079&sdata=QwaxAMupO7ehwbSmN9I%2FFKwgNNr%2FP0th2kuQW6EHb9U%3D&reserved=0> Ireland <https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fdonuts.domains&data=02%7C01%7Cmarksv%40microsoft.com%7Ccd7ec2e952744fc76eb608d68948bdc7%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636847342852603079&sdata=QwaxAMupO7ehwbSmN9I%2FFKwgNNr%2FP0th2kuQW6EHb9U%3D&reserved=0>
Please NOTE: This electronic message, including any attachments, may include privileged, confidential and/or inside information owned by Donuts Inc. . Any distribution or use of this communication by anyone other than the intended recipient(s) is strictly prohibited and may be unlawful. If you are not the intended recipient, please notify the sender by replying to this message and then delete it from your system. Thank you. <https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fdonuts.domains&data=02%7C01%7Cmarksv%40microsoft.com%7Ccd7ec2e952744fc76eb608d68948bdc7%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636847342852603079&sdata=QwaxAMupO7ehwbSmN9I%2FFKwgNNr%2FP0th2kuQW6EHb9U%3D&reserved=0>
On Fri, Feb 1, 2019 at 12:57 AM Alex Deacon <alex@colevalleyconsulting.com> wrote: <https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fdonuts.domains&data=02%7C01%7Cmarksv%40microsoft.com%7Ccd7ec2e952744fc76eb608d68948bdc7%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636847342852603079&sdata=QwaxAMupO7ehwbSmN9I%2FFKwgNNr%2FP0th2kuQW6EHb9U%3D&reserved=0>
Our review of ICANN's input on Temp Spec topics that were not covered by the Initial Report reminded me that we had at one point discussed ensuring that the current Temp Spec language on how Privacy/Proxy data should be handled (Appendix A 2.6) should added as a recommendation. Something along the lines of - <https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fdonuts.domains&data=02%7C01%7Cmarksv%40microsoft.com%7Ccd7ec2e952744fc76eb608d68948bdc7%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636847342852603079&sdata=QwaxAMupO7ehwbSmN9I%2FFKwgNNr%2FP0th2kuQW6EHb9U%3D&reserved=0>
In the case of a domain name registration where a privacy/proxy service used (e.g. where data associated with a natural person is masked), Registrar MUST include in the public WHOIS and return in response to any query full WHOIS data, including the existing privacy/proxy pseudonymized email. <https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fdonuts.domains&data=02%7C01%7Cmarksv%40microsoft.com%7Ccd7ec2e952744fc76eb608d68948bdc7%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636847342852603079&sdata=QwaxAMupO7ehwbSmN9I%2FFKwgNNr%2FP0th2kuQW6EHb9U%3D&reserved=0>
There are two reasons why this is useful, IMO. <https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fdonuts.domains&data=02%7C01%7Cmarksv%40microsoft.com%7Ccd7ec2e952744fc76eb608d68948bdc7%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636847342852603079&sdata=QwaxAMupO7ehwbSmN9I%2FFKwgNNr%2FP0th2kuQW6EHb9U%3D&reserved=0>
First, given the time and effort needed to properly process "reasonable disclosure" requests by Registrars it seems useful to avoid a situation where non-public data is quickly found to be P/P service data. Avoiding this situation and simply including P/P data in the the initial response would make life better for all involved. <https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fdonuts.domains&data=02%7C01%7Cmarksv%40microsoft.com%7Ccd7ec2e952744fc76eb608d68948bdc7%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636847342852603079&sdata=QwaxAMupO7ehwbSmN9I%2FFKwgNNr%2FP0th2kuQW6EHb9U%3D&reserved=0>
Second, there is no need to redact information that is already "redacted" (by definition) by the P/P service. Also, given P/P services list the information of a legal person (in the case of a registrar affiliated service provider) in the place of the RNH's info it seems further redaction is unnecessary. <https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fdonuts.domains&data=02%7C01%7Cmarksv%40microsoft.com%7Ccd7ec2e952744fc76eb608d68948bdc7%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636847342852603079&sdata=QwaxAMupO7ehwbSmN9I%2FFKwgNNr%2FP0th2kuQW6EHb9U%3D&reserved=0>
Happy to discuss further on a future call. <https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fdonuts.domains&data=02%7C01%7Cmarksv%40microsoft.com%7Ccd7ec2e952744fc76eb608d68948bdc7%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636847342852603079&sdata=QwaxAMupO7ehwbSmN9I%2FFKwgNNr%2FP0th2kuQW6EHb9U%3D&reserved=0>
+1.415.488.6009 <https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fdonuts.domains&data=02%7C01%7Cmarksv%40microsoft.com%7Ccd7ec2e952744fc76eb608d68948bdc7%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C636847342852603079&sdata=QwaxAMupO7ehwbSmN9I%2FFKwgNNr%2FP0th2kuQW6EHb9U%3D&reserved=0> _______________________________________________ Gnso-epdp-team mailing list Gnso-epdp-team@icann.org https://mm.icann.org/mailman/listinfo/gnso-epdp-team
_______________________________________________ Gnso-epdp-team mailing list Gnso-epdp-team@icann.org https://mm.icann.org/mailman/listinfo/gnso-epdp-team
_______________________________________________ Gnso-epdp-team mailing list Gnso-epdp-team@icann.org https://mm.icann.org/mailman/listinfo/gnso-epdp-team