Eve, can you unpack the technical solution point that you're making? Are you saying that an n:1 solution is currently incompatible with a 1:1 solution and which role is on either side of the : ?

Adrian

On Mon, Jan 25, 2016 at 3:23 PM, Eve Maler <eve.maler@forgerock.com> wrote:
I'm very glad to see this clear elucidation of where the splits in understanding/belief are.

The UMA WG's design center never actually assumed or required a strict 1:1 relationship between RO and AS, as can be seen in the charter and requirements and design principles.

The nature of agency law, as I understand it, is precisely to ensure that the agent's interests align properly with those of the principal so that the agent acts within their authority when they act on behalf of the principal. This should hold true whether there is a 1:1 or n:1 relationship, and it's part of why the UMA legal subgroup is doing its work.

Of course, if technology can do a good job of keeping an agent in line, then a legal remedy might not be required. But so far, I haven't seen jwk be a good candidate for a technical solution.

People have asked me about various encryption or DRM solutions being used on top of UMA. Generally my response to them is that they can use such a thing if they want to. However, solutions involving encryption can impose costs that many ecosystems can't bear (particularly wide ecosystems with lots of third-party clients -- witness OAuth V1.0), and often the players end up treating such requirements as failure and route around them by emailing content or sharing passwords. :-) Access control systems must make the right thing to do be the easiest thing to do.

Eve Maler
ForgeRock Office of the CTO | VP Innovation & Emerging Technology
Cell +1 425.345.6756 | Skype: xmlgrrl | Twitter: @xmlgrrl
Join our ForgeRock.org OpenUMA community!


On Mon, Jan 25, 2016 at 9:32 AM, Adrian Gropper <agropper@healthurl.com> wrote:
(apologies for cross-posting)

Thanks, Josh. As always, you are able to explain the issue much better than me. This is why I've tried my best to understand the gaps between UMA as it is today and something that lays cliaim to user-managed access. FHIR and UMA and, consequently, HEART are all new and we need to be very clear about how we design this next generation of protocols.

In the UMA legal subgroup, we've tried to map UMA into "Agency Law", the relatively simple branch of law that dictates how an individual (patient) can specify an agent (as in lawyer, accountant, or doctor) to act on their behalf. Suffice it to say, that agency law does not assume that an agent will have any conflict of interest in the baseline case but that conflicts of interest can arise in the real world (for example when a broker is put in an agency role).

It's up to all of us, FHIR, UMA, and HEART, to develop around the requirement for agency on the part of the individual. We can layer on all sorts of other complexities to deal with brokerage and jurisdictional issues, but if we don't start with clear personal agency in the Authorization Server, then we will keep chasing the increased complexity and insecurity of our networks for another 10 years.

Adrian


On Monday, January 25, 2016, Josh Mandel <Joshua.Mandel@childrens.harvard.edu> wrote:
That's illuminating, at least. It sounds like you're interested in changing the underlying protocol (i.e. effectively designing something new instead of UMA). There's nothing wrong with that, but making these changes implicitly and then calling the new thing "UMA" is exacerbating the confusion on this list.

Also, I can't help responding to the following comment that you made on #5:
You're mixing up the identity of an authorization service with the identity of a user or persona.

 -- that's exactly correct! Using the AS's public key to identify the user would be mixing up the identity of the AS with the user. That's exactly why I'm saying you *cannot equate* them. So perhaps we agree on this point at least :-)

  -J 

On Mon, Jan 25, 2016 at 11:56 AM, Adrian Gropper <agropper@healthurl.com> wrote:


On Mon, Jan 25, 2016 at 11:49 AM, Josh Mandel <Joshua.Mandel@childrens.harvard.edu> wrote:
Adrian, I'm afraid we're talking past one another here. Let me try to spell out the logic and see if you follow and agree with each step:

1. UMA allows a user to stand up her own Authorization Server
Yes
2. UMA does not require a user to stand up her own Authorization Server
I don't know. It may be that's the only way for an UMA-like approach to privacy and security to scale.
3. Given #2, an UMA-based protocol cannot assume that every user will stand up her own Authorization Server
Why not? Can't a service host an arbitrary number of AS?
4. Given #3, an UMA-based protocol must assume that some Authorization Servers will work on behalf of multiple users
That depends on how we choose to define Authorization Server.
5. Given #4, an UMA-based protocol cannot equate an Authorization Server's identity with a user's identity
You're mixing up the identity of an authorization service with the identity of a user or persona.
6. Given #5, an UMA-based protocol cannot use an Authorization Servers public key to identify a user.
It's all in how we define Authorization Server, isn't it?

Adrian

Is there a point at which you disagree?

  -J 

On Mon, Jan 25, 2016 at 11:40 AM, Adrian Gropper <agropper@healthurl.com> wrote:
No, I'm not saying anything about how the public key associated with a persona is used for encrypting messages to the client. The use of a public key the way to identify a persona or account is already well established in blockchain and is completely compatible with a personal owned AS. That's all I'm saying.

The use of whatever keys or tokens will be used in messaging between the RS and the Client is up to the "security folk" Justin refers to. Are the security folk saying that having a personal persona AS is impossible?

Adrian

On Mon, Jan 25, 2016 at 11:34 AM, Josh Mandel <Joshua.Mandel@childrens.harvard.edu> wrote:
In UMA, the RS does indeed learn the AS's public key (by asking the AS). But the public key is not used in the way you seem to want (namely, it's not used to encrypt any patient data). It sounds to me like you're focusing on an implementation detail (the fact that the AS happens to have a public key associated with it), and extrapolating to some unfounded conclusions (that this forms the basis for encrypting patient data with a patient-specific key) from that detail.

  -J

On Mon, Jan 25, 2016 at 11:26 AM, Adrian Gropper <agropper@healthurl.com> wrote:
Justin,

All I'm saying is that each patient persona gets to have a private key and that key is kept safe in their AS - period.

Let's start with that, and then the security folks can tell us how the RS gets the Client's public key. Can't UMA do that?

Adrian

On Mon, Jan 25, 2016 at 11:02 AM, Justin Richer <jricher@mit.edu> wrote:
The way encryption is handled between the RS and the Client does not follow from the right to choose an AS and has nothing to do with the AS’s JWK. Think about it this way: how would the client get the AS’s private key to decrypt the message? 

 — Justin

On Jan 25, 2016, at 10:52 AM, Adrian Gropper <agropper@healthurl.com> wrote:

I think we're mostly all on the same wavelength. The first priority is to bake the right to control one's own AS and the corresponding jwk into FHIR and HEART. This should be easy to reach consensus on given recent OCR guidance that includes the "right to delegate access to a third party".

The way encryption is handled between the RS and Client follows from the above.

Adrian

On Mon, Jan 25, 2016 at 10:38 AM, Josh Mandel <Joshua.Mandel@childrens.harvard.edu> wrote:
Let's generously read Adrian's claim here as "a patient who cares enough can set up her own AS, and thereby ensure that her AS's jwk is unique to her". So yes, you can (with enough effort -- which might be driven down by technology) get to the point of one jwk per AS, if you really want.

But even if you do so, the AS's jwk in HEART isn't really an "encryption key" sense in the way Adrian would have it. In other words, the AS's jwk is not used as an "encryption key" for the patient's healthcare data at any point. (It is used to sign tokens in the UMA flow, including the PAT and the AAT.)

By the way, http://openid.bitbucket.org/HEART/openid-heart-uma.html#rfc.section.2.1 says that an "aud" claim in the AAT specifies the "RPT authorization endpoint"; UMA refers to this as simply the "RPT endpoint", and I think the HEART profile's language should be consistent (this tripped me up for 10min just now).

  -J



On Mon, Jan 25, 2016 at 10:02 AM, Adrian Gropper <agropper@healthurl.com> wrote:
Why "most likely not"? Is it a security issue? a cost issue? We don't have to compromise privacy for security in our connected world.

On Mon, Jan 25, 2016 at 9:55 AM, Justin Richer <jricher@mit.edu> wrote:
But it's not like that, the arity is very different.

Every record is associated with an AS, perhaps a separate AS for each record/patient but most likely not.

Every AS is associated with a jwks_uri, but only one per AS.

 -- Justin


On 1/25/2016 9:02 AM, Adrian Gropper wrote:
It means that every patient record is associated with a separate jwks_uri for that patient's AS.

On Mon, Jan 25, 2016 at 8:59 AM, Justin Richer <jricher@mit.edu> wrote:
Yes you did. Quote:

"The system is also much more resistant to data breaches as data holders (UMA Resource Servers) must implement separate encryption keys for each patient."

 So if you don't mean separately encrypting the data for each user, what does that statement mean? The access token isn't an encryption key.

 -- Justin


On 1/25/2016 8:57 AM, Adrian Gropper wrote:
I never said anything about how the data is encrypted. I only talk about how access to the FHIR API is controlled.

Adrian

On Mon, Jan 25, 2016 at 8:55 AM, Justin Richer <jricher@mit.edu> wrote:
Adrian,

I've asked this before and thought we'd settled it, but it keeps coming up: where are you getting the idea of encrypting the data to the patient using a patient's key? That is not in scope for HEART, nor is it part of any of the underlying protocols.

 -- Justin


On 1/25/2016 8:52 AM, Adrian Gropper wrote:
Establishing a separate URI for each patient is likely to be the only stable solution to the patient ID problem. The issue, however, is how many URIs will a patient be allowed to have? If the URIs are coercive, in the sense of a chip or tattoo issued by government or an equivalent global authority (Facebook?) or the URI is derived from DNA or an iris scan. (Iris scans are a good positive IDs and can be read from 30 feet away with modern technology.)

Let's assume, for our purposes, that an iris scanner costs about as much as a credit card terminal, cheap enough for every front office, ambulance, and police car. Is the patient ID problem solved? I don't think so.

Patients can have one or more separate URIs in order to help manage their health records. Today, we typically use email address for this purpose, with WebFinger https://webfinger.net/ as a standardized way to discover linked attributes such as the patient's UMA Authorization Server and the associated public key.

UMA for patient ID brings numerous benefits including much greater transparency and security. The patient now has a single portal (their UMA AS) to view all current relationships under that particular patient ID persona. The system is also much more resistant to data breaches as data holders (UMA Resource Servers) must implement separate encryption keys for each patient.

I think the HEART group is in a good position to compete for the CHIME challenge on this basis and I'd be happy for me and PPR to help organize a submission.

Adrian

On Sun, Jan 24, 2016 at 1:20 PM, Aaron Seib <aaron.seib@nate-trust.org> wrote:
I appreciate your expertise and action.  

I don't necessarily agree with some of your statements here but that is the beauty of open processes.  

Let's strive to do all we can - together.



Aaron Seib
@CaptBlueButton

"The trick to earning trust is to avoid all tricks.  Including tricks on yourself."



-------- Original message --------
From: "Glen Marshall [SRS]" <gfm@securityrs.com>
Date: 2016/01/24 7:07 AM (GMT-08:00)
To: HEART List <openid-specs-heart@lists.openid.net>
Subject: [Openid-specs-heart] CHIME Launches $1M Challenge to Solve Patient ID Problem

This is pertinent to our data-sharing use cases.  There is no current solution to accurately sharing/gathering patients' clinical data stored among various repositories.  In turn, that makes applying access controls across all of a patient's data in those repositories difficult.   I'm happy to see Chime's challenge.

However, the related problem of discovering where all of one's data might be is computationally intractable.  It is equally intractable to gather and combine all access permissions and regulatory restrictions on patients' data, even if there were a useful means to do so.  (Both are equivalent to the halting problem.)

Having a single "source of truth" repository is one direction for a solution, as is having a single access permissions source.  Keeping them updated with new data and permissions is possible, even if difficult in the short run.

However, establishing unique URIs for each patient's data and permissions is the same as having a universal patient identifier.  That might be subject to current Congressional funding restrictions.  


The College of Healthcare Information Management Executives on Tuesday launched a $1 million National Patient ID Challenge to develop solutions to ensure 100 percent accuracy of every patient’s identity to reduce preventable medical errors.

http://www.healthdatamanagement.com/news/chime-launches-1m-challenge-to-solve-patient-id-problem
--

Glen F. Marshall
Consultant
Security Risk Solutions, Inc.
698 Fishermans Bend
Mount Pleasant, SC 29464
Tel: (610) 644-2452
Mobile: (610) 613-3084
gfm@securityrs.com
www.SecurityRiskSolutions.com


_______________________________________________
Openid-specs-heart mailing list
Openid-specs-heart@lists.openid.net
http://lists.openid.net/mailman/listinfo/openid-specs-heart




--

Adrian Gropper MD

PROTECT YOUR FUTURE - RESTORE Health Privacy!
HELP us fight for the right to control personal health data.
DONATE: http://patientprivacyrights.org/donate-2/ 


_______________________________________________
Openid-specs-heart mailing list
Openid-specs-heart@lists.openid.net
http://lists.openid.net/mailman/listinfo/openid-specs-heart




--

Adrian Gropper MD

PROTECT YOUR FUTURE - RESTORE Health Privacy!
HELP us fight for the right to control personal health data.
DONATE: http://patientprivacyrights.org/donate-2/




--

Adrian Gropper MD

PROTECT YOUR FUTURE - RESTORE Health Privacy!
HELP us fight for the right to control personal health data.
DONATE: http://patientprivacyrights.org/donate-2/




--

Adrian Gropper MD

PROTECT YOUR FUTURE - RESTORE Health Privacy!
HELP us fight for the right to control personal health data.
DONATE: http://patientprivacyrights.org/donate-2/

_______________________________________________
Openid-specs-heart mailing list
Openid-specs-heart@lists.openid.net
https://urldefense.proofpoint.com/v2/url?u=http-3A__lists.openid.net_mailman_listinfo_openid-2Dspecs-2Dheart&d=BQICAg&c=qS4goWBT7poplM69zy_3xhKwEW14JZMSdioCoppxeFU&r=c7b1QeR755-dBx2b0xnlepDTylromoEzcLl-6ixmBL3TpXSxSvtAvT553fzSgLpm&m=2VMieJq3EmwTLglsueFRPjnL6SoON964_PI0l93fF_Q&s=YlZbL57zFJCHlm8lTsjPaao9DWipNm9s50Q8ZFOgk54&e=






--

Adrian Gropper MD

PROTECT YOUR FUTURE - RESTORE Health Privacy!
HELP us fight for the right to control personal health data.
DONATE: http://patientprivacyrights.org/donate-2/




--

Adrian Gropper MD

PROTECT YOUR FUTURE - RESTORE Health Privacy!
HELP us fight for the right to control personal health data.
DONATE: http://patientprivacyrights.org/donate-2/




--

Adrian Gropper MD

PROTECT YOUR FUTURE - RESTORE Health Privacy!
HELP us fight for the right to control personal health data.
DONATE: http://patientprivacyrights.org/donate-2/




--

Adrian Gropper MD

PROTECT YOUR FUTURE - RESTORE Health Privacy!
HELP us fight for the right to control personal health data.
DONATE: http://patientprivacyrights.org/donate-2/



--

Adrian Gropper MD

PROTECT YOUR FUTURE - RESTORE Health Privacy!
HELP us fight for the right to control personal health data.
DONATE: http://patientprivacyrights.org/donate-2/


_______________________________________________
Openid-specs-heart mailing list
Openid-specs-heart@lists.openid.net
http://lists.openid.net/mailman/listinfo/openid-specs-heart





--

Adrian Gropper MD

PROTECT YOUR FUTURE - RESTORE Health Privacy!
HELP us fight for the right to control personal health data.
DONATE: http://patientprivacyrights.org/donate-2/