OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.

 


Help: OASIS Mailing Lists Help | MarkMail Help

soa-rm-ra message

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]


Subject: for today's call [was: [soa-rm-ra] Re: [soa-rm] status and near term activities for SOA-RAF]


What I get out of this exchange is Michael is saying certain things don’t work and Kevin saying there are standards and solutions in use that support these nonworking things and are not going away.  Maybe that is just the amateur’s take but it is also my general observation.

 

So how do we proceed?  I see the need for the following:

-          What are security challenges to be addressed and, in particular, what are different in a SOA ecosystem?

-          What are known approaches to dealing with those challenges? Again, how do those approaches change/need to change/been observed to change for SOA ecosystem?

-          What do those approaches actually provide and what are their weaknesses?

-          Given an incomplete solution, what does our audience need to keep in mind, what questions do they need to ask?

 

We should not be taking sides in a debate but rather should be providing solid information for the debaters.

 

I am late for the call, but let’s use this for discussion.

 

Ken

 

---------------------------------------------------------------------------

Dr. Kenneth Laskey

MITRE Corporation, M/S H305              phone: 703-983-7934

7515 Colshire Drive                                    fax:        703-983-1379

McLean VA 22102-7508

 

From: soa-rm-ra@lists.oasis-open.org [mailto:soa-rm-ra@lists.oasis-open.org] On Behalf Of Kevin Smith
Sent: Wednesday, July 25, 2012 10:22 AM
To: Mike Poulin
Cc: soa-rm-ra@lists.oasis-open.org
Subject: Re: [soa-rm-ra] Re: [soa-rm] status and near term activities for SOA-RAF

 

Mike,

I apologize that I missed the call last week - I was in a meeting that overlapped.

I have been thinking about your comments this week and how exactly to respond.

In the community that I work in, it is extremely important for services to be able to provide access control to data based on authorization credentials of the end-user.  There are organizations that require that data that each service accesses is filtered based on credentials of the ultimate end-user (and I know this is where you disagree).  

In a service orchestration scenario, this is achieved by token passing related to identity/pseudo-identity/authorization credentials/authorization statements. (I won't limit this to identity). There are quite a few models for doing this done by quite a few standards. Many of these standards also presuppose & are based on certain identity infrastructure. This identity infrastructure provides context related to the token that is passed. New standards, as we have discussed, provide "delegated consent" - meaning that a user delegates the ability to act on behalf of that user to a service (so the service can then tell another service that it is acting on behalf of a user).

In the model you describe ("A Consumer of my Consumer is not my Consumer"), you seem to imply that each service would only provide access control based on its consumer. In a service orchestration scenario, this would leave the final burden of the filtering of the data on the end-user's application (as the user's application is the user's consumer). At this point, unless the data that is returned by the application's service is marked with security tags, the application may not have enough information to filter data based on the user and his authorization credentials.  Not only may there be problems with this model, but the model you suggest may not meet the security requirements of many organizations.

Certainly, many people are wrestling with these issues & standards continue to progress.  But token propagation (I won't limit this to identity) is a practical reality.

I guess my biggest concern is that the Security Section of this document has been written and static for some time - I confess that I am not certain what to do with all of your suggestions at this point!

-Kevin

On Wed, Jul 25, 2012 at 9:41 AM, Mike Poulin <mpoulin@usa.com> wrote:

Ken,

here are my responses.

re "In general, we have not explicitly referenced other standards"  - security is standard-rich area as well as WS-* and if we try to _describe_ something in security, it might look like a new standard definition for security. I do not think we are in this, security business. This is why I propose to refer instead of describe or, at least, quote stnadard definitions of all core security aspects/concerns (integrity, confidentiality, etc.).


re “is replaced by” -  I agree with your words. What I said in the comment to the topic is an implementation consequence: "Centralized security management mechanism is replaced by a chain of Service Contracts" of your words. This is one of the most unexpected and dameging results for security caused by ownership service boundaries. Yes, depending on policies, these boundaries my be invisible, but what to do if they are visible? Actually, you said the same as I did but you used "policies" while I used "Service Contracts" that contain policies.


re “A model of Identity Federation replaces a model of Federated Identity -  you will not see a concensus soon becuase majority of corporate security and vendors promote Federated Identity (for SSO, for example) which does not work in Clouds and recognised (back to 2010) as broken by several authors. In short, Federated Identity model assue that an identity of a user is carried by a token that can be recognized in several security/authentication realms (I omit internal structure of the token). The _only_ condition under which Federated Identity can work is an _agreement_ of all authentication realms to recognise this token. Becuase of ownership boundaries and commercial independence of Cloud Providers, reaching such an agreement is highly questionable matter. An Identity Federation is a federation of security/authentication realms that do not agree all together on anything but in each pair have an agreemnt and a mechanism how to propagate a request from each other under their individula authentication controls.

I have attached a draft of my article (currently in the editor process in InfoQ) regarding this topic.


RE "some of your conclusions are not universally accepted" -  well, SOA Ecosystem is also a not universally accepted
concept... "but allow room for other solutions to be developed in the future. A significant challenge is to demonstrate unwanted consequences of some brute force approaches that don’t scale." -  quite agree.


 

----- Original Message -----

From: Ken Laskey

Sent: 07/25/12 01:48 PM

To: 'Mike Poulin', soa-rm-ra@lists.oasis-open.org

Subject: RE: [soa-rm-ra] Re: [soa-rm] status and near term activities for SOA-RAF

 

Michael,

 

 

 

 

 

Thanks for the security topic list.  Sorry for the delay in responding but I was on leave most of last week and still dealing with accumulated email.

 

 

 

 

 

Here are my thoughts on your topics:

 

 

1.       re “with references to related security standards”: In general, we have not explicitly referenced other standards unless, for example IEEE-1471, we used these in developing the RAF. The limited exception is footnotes for obvious connections, such as footnote 6 on p. 52 of the 6 July draft.

 

 

2.       re “is replaced by” in section 1.1.2.1: This is an example where I believe we need to be less prescriptive. At the RAF level of abstraction, we need to say that a consequence of chaining services where those services can be in different ownership domains is that conditions of use (e.g. security policies) as captured in the service contract may vary from domain to domain and this implies some level of policy enforcement is required for each service. This may be supported by services providing security functions along with local enforcement.

 

 

3.       re “A model of Identity Federation replaces a model of Federated Identity”: Explain differences and consequences of each. This should lead to conclusions of what to use but does not take a stand on the “right way”.  I’ve seen a lot of discussion but not consensus.

 

 

4.       I agree with many of the points you make in the 1.1.2.x sections but I also know that some of your conclusions are not universally accepted.  We need to show how the SOA paradigm leads to situations where one could reach these conclusions but allow room for other solutions to be developed in the future. A significant challenge is to demonstrate unwanted consequences of some brute force approaches that don’t scale.

 

 

5.       re Security Responses: Agree that we need to be clear what is standard techniques that are relevant to SOA vs. where a SOA implementation introduces something new.

 

 

 

 

 

Ken

 

 

 

 

 

---------------------------------------------------------------------------

 

 

Dr. Kenneth Laskey

 

 

MITRE Corporation, M/S H305              phone: 703-983-7934

 

 

7515 Colshire Drive                                    fax:        703-983-1379

 

 

McLean VA 22102-7508

 

 

 

 

 

From: soa-rm-ra@lists.oasis-open.org [mailto:soa-rm-ra@lists.oasis-open.org] On Behalf Of Mike Poulin
Sent: Sunday, July 22, 2012 8:43 AM
To: soa-rm-ra@lists.oasis-open.org
Subject: [soa-rm-ra] Re: [soa-rm] status and near term activities for SOA-RAF

 

 

 

 

 

Hi All,

as we agreed in the last telecom, I offer a list of topics and a few comments for the Security Model section. I am interested in your version of this list.


 

 

 

Security Model

 

Definition of security and preamble

 

 

1.1.1 Secure Interaction Concepts

 

Briefly, with references to related security standards:

 

 

1.1.1.1 Confidentiality

 

1.1.1.2 Integrity

 

1.1.1.3 Authentication

 

1.1.1.4 Authorization

 

1.1.1.5 Non-repudiation

 

1.1.1.6 Availability

 

1.1.1.7 Inter-relationship of Security Concepts (diagram

 

1.1.2 SOA Specific Approaches to Security Concerns

 

service independence à service ownership boundaries à business/enterprise boundaries à security realm boundaries à security realm federation

 

 

1.1.2.1 Security Concepts in Service Conract

 

Centralized security management mechanism is replaced by a chain of Service Contracts.

 

 

Individual  Service Contracts  define (or refer to) all policies for identity and access controls as well as for confidentiality, integrity and non-repudiation policies. In SOA Ecosystem, policy enforcement for all listed security concepts is based on individual Service Contracts.

 

 

A centralized security management may still be preserved in a limited group of participants who know each other and agree to open their service-business-security boundaries based on mutual trust and shared interests.

 

 

1.1.2.2 Service Ownership Boundaries

 

Consequences of Service Ownership Boundaries:

 

 

·         ·         A Service of my Service, is not my Service

 

 

·         ·         A Consumer of my Consumer, is not my Consumer

 

 

·         ·         A Partner of my Partner, is not my Partner

 

 

·         ·         A Supplier of my Supplier, is not my Supplier

 

 

 

 

 

A model of Identity Federation replaces a model of Federated Identity. Identity Federation establishes a mechanism of trust via intermediaries shared between federated security relams above the level of consumers and services but for them.

 

 

1.1.2.3 Authentication and Ownership

 

Because of ownership boundaries in SOA Ecosystem, authentication is required only in the scope of service interaction. A propagation of an identity of an initial requester through the chain of interacting services is useless and wasting resources.

 

 

If a service consumer interacts with the service because being engaged by its own consumer, the service reponses to the service consumer based on their Service Contract regardless the reasons of this interaction.

 

 

1.1.2.4 Authorization and Ownership

 

Because of ownership boundaries in SOA Ecosystem, authorization is required only in the scope of service interaction. A consumer does not have any access rights to the service and cannot acquire them in any other way than via Service Contract. An implementation mechanism of granting agreed access rights may vary.

 

 

A propagation of access rights of an initial requester through the chain of interacting services is useless and constitutes a confidentiality violation.

 

 

1.1.2.5 Confidentiality and Ownership

 

Each security realm may have its own confidentiality policies for service communication and execution within individual ownership boundaries and for the individual service interaction case. Particular policies are regulated by an individual Service Contract.

 

 

1.1.2.6 Integrity and Ownership

 

Each security realm may have its own integrity policies for service communication within individual ownership boundaries and for and for the individual service interaction case. Particular policies are regulated by an individual Service Contract.

 

 

1.1.2.7 Non-repudiation and Ownership

 

Each service may establish its own policies for non-repudiation for its consumers. One service is not obliged to preserve non-repudiation policies of its own  consumer or another service if this obligation is not specified in the Service Contract.

 

 

1.1.3 Security Threats for Services

 

Enumeration of threads with no explanations. Only  types of threads specific to SOA services should be elaborated. [A Web specific threat like Denile Attack/Replay Attack or common threats like Message Alteration, Message Interception, Spoofing, Man in the middle – they are not SOA specific]. IMO, only False Repudiation has specifics for SOA because of Service Contract.

 

 

- 1.1.4 Security Responses – not SOA specific and not needed

 

1.1.5 Architectural Implications of SOA Security

 

as appropriate

 

 

 



---------------------------------------------------------------------
To unsubscribe, e-mail: soa-rm-ra-unsubscribe@lists.oasis-open.org
For additional commands, e-mail: soa-rm-ra-help@lists.oasis-open.org

 



[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]