|Moehrke, John (GE Healthcare)||Jul 14, 2009 10:07 am|
|Davis, John M.||Jul 14, 2009 11:39 am|
|Moehrke, John (GE Healthcare)||Jul 15, 2009 4:03 am|
|Anthony Mallia||Jul 15, 2009 7:11 am|
|Moehrke, John (GE Healthcare)||Jul 15, 2009 2:45 pm|
|Anthony Mallia||Jul 16, 2009 6:46 am|
|Moehrke, John (GE Healthcare)||Jul 16, 2009 12:18 pm|
|Moehrke, John (GE Healthcare)||Jul 16, 2009 12:22 pm|
|steven meyer||Jul 16, 2009 2:01 pm|
|Subject:||RE: [xspa] Policy scope|
|From:||Moehrke, John (GE Healthcare) (John...@med.ge.com)|
|Date:||Jul 16, 2009 12:22:42 pm|
I should have added... That this issue is explicitly a problem with Cross-Enterprise (The X in XSPA)... So, it is possible XACML core doesn't have a solution, and that XSPA is the right place to solve the problem..
From: Moehrke, John (GE Healthcare) Sent: Thursday, July 16, 2009 2:19 PM To: Anthony Mallia; Davis, John M.; xs...@lists.oasis-open.org Subject: RE: [xspa] Policy scope
Yes, this is the topic I am interested in... How would the XACML policy look like that shows which parties, or what the policy domain is...
From: Anthony Mallia [mailto:amal...@edmondsci.com] Sent: Thursday, July 16, 2009 8:50 AM To: Moehrke, John (GE Healthcare); Davis, John M.; xs...@lists.oasis-open.org Subject: RE: [xspa] Policy scope
In these discussions there appears to be a more basic underlying theme which we have not directly addressed but comes up when we discuss specific solutions (e.g. NHIN). Policy scope could be thought as policy domain which is the environment for policy formulation, acceptance and possibly its enforcement.
A number of policy domains already exist - federal and state jurisdiction, enterprise or covered entity. An organization may be required to operate within multiple policy domains and a couple of approaches can be taken - consolidation or isolation. If a healthcare provider is operating in multiple states it is faced with these choices. Sometimes isolation is not possible.
I believe that both approaches are valid and that the solutions should facilitate them both.
When we consider the patient consent directive (a type of policy), US law specifies that it is the covered entity which may accept that policy and if so must enforce it. There is at the moment no higher level policy domain which can accept and enforce that policy. However, by mutual agreement between covered entities, the patient might be able to make the formulation and have it accepted by multiple covered entities thereby simplifying the complexity for the patient. This appears to be a main driver in these discussions due to the autonomy of the covered entity.
Hopefully this starts to address your last sentence which reflects the need to identify who has accepted the policy since another party is under no legal obligation to enforce it. Maybe we need a language for describing what a colleague has described as "Policy Engineering". This would cover the concepts in law such as restriction, authorization (HIPAA Privacy) and the lifecycle of formulation, acceptance and enforcement. XACML could refer to this language either directly in naming or indirectly through comments.
I agree with what you say. And it is your last paragraph that is the focus of my question.
So, to the general XACML audience I am interested in how this might be enabled by general XACML standard. If it is not, then I suggest that XSPA will need to address it as it is related to cross community security and privacy authorization. Healthcare does have a specific need, yes. I just don't think this is unique to healthcare, nor NHIN as the USA example.
From: Anthony Mallia [mailto:amal...@edmondsci.com] Sent: Wednesday, July 15, 2009 9:15 AM To: Moehrke, John (GE Healthcare); Davis, John M.; xs...@lists.oasis-open.org Subject: RE: [xspa] Policy scope
A small point has been missed in the discussion. The NHIN DURSA autonomy principle allows that the requesting community and the responding community may have totally separate policies. They do not even need to know each other's policies.
The requesting community has an ACS which it uses to authorize the request based on its own policies (institutional and consumer preferences) and the responder has its ACS where it authorizes the response based on its own policies. In order for the complete transaction to be successful, both must authorize.
Between the two communities, the NHIN Authorization Framework carries the assertions which were used by the requesting community to authorize the request and this is equivalent to the XSPA SAML profile where the firsts community is making a request to the second community. The conformance point is clearly in scope for the NHIN.
The PEP/PDP exchange is within each ACS and is not exposed to the NHIN and therefore is out of scope for NHIN but not for OASIS. The XSPA XACML Profile conformance point (which I believe is between the PEP and the PDP) is important because it gets many of its attributes (but not all) from the XSPA SAML Profile.
From the view of harmonizing NHIN and XSPA we only need to align the XSPA SAML Profile and the Authorization Framework.
The second discussion which is the exchange of consumer preferences or consent directives is a broader and more complex topic. However it must be harmonized with the XSPA XACML profile which provides attributes for the PDP decision.
Edmond Scientific Company
4100 Monument Corner Drive, Suite 540 Tel: (703) 766-3871, Ext. 400
Fairfax, Virginia 22030-8610 Fax: (703) 766-3872
Mike, I am glad you expressed how this is handled in the NHIN DURSA world. This is also a good example of how the NHIN DURSA got around the issue. This solution is based on the tools they had available at the time. What I am trying to get discussed is if there is a more flexible way to support this in the XSPA standards future.
The problem I see with the NHIN DURSA solution is that it simply makes the problem the responsibility of ACS requestor. This does not eliminate the problem, it simply moves the problem out of DURSA space and into the hie/hospital/clinic space where, in general, there is little expertise to solve the problem. I am well aware that you personally are happy with this solution as you do have solutions and architects in the VA space. I think we must think of the whole architecture. This is not to say that we must define an internal architecture for the 'participants', but the underlying XACML schema seems should support appropriate metadata coding to support the need.
As you very clearly indicated a policy is not enforced until both parties agree to enforce it. So, how is this 'scope' of the policy encoded? This is my question.
So to be clear. The reason I am asking this question on XSPA is because I see this as an unsolved problem that our 'standards' development should consider. I am well aware of ways that it is handled in the old days, as I also indicated one of these old days as well.
The NHIN DURSA follows the principal of "autonomy", meaning that each participant follows it's own policies in making access control decisions. The NHIN does not gather policies, the participants do, but only from their own policy stores. When a request is made, it is decided upon based upon the participants policy and the consent directives on hand.
In the case of the VA, for example, we would not release information unless there was a consent directive to do so (Opt-In is a minimal consent directive). That means that anyone requesting information must provide the consent directive at the time of the request. There are many ways to do this, for example, by reference to an XDS directory.
Date reuse requires that the patient establish appropriate consent directives at each participant site. The patient consent directive at the responding participant needs to reflect their preferences established with the requesting participant.
Again, the autonomy principal requires that each participant have all policies on hand for their decision, since they are the authoritative source for all policies, including patient ones that they are enforcing. A participant is NOT responsible for enforcing policies not in its policy repository. BTW, Consent directives themselves are not automatically enforced and do not become participant policies until the agency agrees to enforce them. This enforcement agreement is a binding contract between the participant and the patient. That requires an administrative action (since it involves liability on the provider) as part of the process. Finally, participants will not push a consent directive as an obligation.
So what is important are the attributes asserted at the time of the request. These are used to inform the participant policies (whatever they are). Since different participants are under jurisdiction of different laws, there can be no guarantee of a "common" set of enterprise policies, but I believe that might be possible with consent directives which will probably be related to specific law at some point. On the other hand, Federal law (e.g., FISMA) that applies only to federal agencies may require different policies than California does. Since we are not passing policies back and forth, this should be transparent. In other words, the requester only needs to know what authorization attributes need to be asserted, not what policy set is being enforced. Also, there is no burden on a participant to tell another "Why" a decision was revoked, only to respond, e.g. "Data not available".
Here are the core policy assumptions:
* Multi-Party Agreement. Single agreement to which all Participants agree that establishes trust framework and rules of engagement.
* Participants in Production. Assumes the agreement is for those in production and complements end user trust agreements.
* Privacy and Security Obligations. Defers to Applicable Law and establishes HIPAA as contractual standard of performance for other non-HIPAA or non-governmental entities.
* Requests for Data Based on Permitted Purposes. Permits exchange of information among NHIN for certain purposes for the pilots, including treatment, payment, health care operations, public health activities, quality reporting for "meaningful use" and disclosures based on an authorization from the individual (e.g. SSA disability determination).
* Duty to Respond. All Participants have a duty to respond to requests for data for treatment purposes, and may respond to other types of requests.
* Future Use of Data Received Through the NHIN. Data are received and integrated into end-user's system and may be reused or disclosed as any other information in its records, in accordance with Applicable Law and local policy.
* Duties of Requesting and Responding Participants. Each Participant has certain duties when acting as a requesting or responding Participant, including conformance with NHIN requirements.
* Autonomy principle for access - When responding to a request for data, Participants will apply their local policies to determine whether and how to respond to the request.
* Responding Participant's Legal requirements. A Responding Participant must make sure that it has met any legal requirements before disclosing the data including, but not limited to, obtaining consent or authorization for treatment purposes.
* Authorizations. When a request is based on an authorization (e.g. for SSA benefits determination), the requesting Participant must send a copy of the authorization with the request for data.
* Participant Breach Notification. Participants are required to promptly notify the NHIN Coordinating Committee and other impacted Participants of breaches which involve the unauthorized disclosure of data through the NHIN.
* Mandatory Non-Binding Dispute Resolution. Participants will agree to participate in a mandatory, non-binding dispute resolution process that preserves the Participants' rights to seek redress in the courts if not resolved through the dispute resolution process.
* Allocation of Liability Risk. With respect to liability, the DURSA memorializes the Participant's understanding that each Participant is responsible for its own acts or omissions.
Regards, Mike Davis, CISSP
Department of Veterans Affairs
CHIO Emerging Health Technologies
A question came up today in discussion on Access Control white paper in IHE. How is it known when all applicable policies have been gathered? Specifically in an environment like the USA NHIN, how does the NHIN know when they have done sufficient gathering of policies in order to make the specific-access-control-decision?
I have heard discussions where the access control decision logic simply is happy making a positive response based on the knowledge it has, but how does it not know that there is a 'new' policy declaration available that indicates that the patient has revoked authorization?
The same can be said for a decision of NO could have been rather a YES because of a new authorization.
But the most concerning is how does the system understand that there is a policy available to be gathered/retrieved that is simply not applicable.
In the 'old days' under BPPC, this is handled simply because each policy identifier is either understood by the access-control-service (meaning the policy logic was configured into the ACS) or it is NOT. Therefore a query for all policies that have been acknowledged results in a list of unique identifiers to enable, therefore the decision is made based on the cross-section of those that are acknowledged vs those that are understood. This results in a self limited set of policies.
In the XACML world, it is possible for the ACS to pull down the computable policy and make it known to my ACS... so how do I determine which I SHOULD incorporate vs which ones should I continue to ignore?
I presume that much of this is simply encoded into the policy as the scope of the policy. For example which organizations does this policy affect. If a policy indicates that it is scoped to a set of organizations that are not involved in a transaction, then that policy is not applicable. Is this already built into XACML core?
John Moehrke Principal Engineer: Interoperability and Security GE Healthcare
M +1 920 912 8451
9900 Innovation Drive
Wauwatosa, WI 53226
GE imagination at work