[data-protocols] Access Control

[Haq, Salman]

Francis,

So let's stick with the API key concept. Suppose that instead of (or in
addition to) each individual having a unique key, they were assigned a
re-usable key from a set of keys that represent roles. Roles can be
arbitrarily designed to have specific privileges. An example would be a
role that can access email address information (role A). Another example
would be a role that can access email addresses and street addresses (role
B).

Now an individual having role A formulates a query that fetches email and
street addresses. The query completes partially, returning only email
addresses because their role is not authorized to access other types of
information.
If the same query had been formulated by an individual with role B, it
would have succeeded completely.

Notice that I am concerned with semantic (functional) data types. The
actual values may be stored in native formats and may even differ from
schema to schema (eg: a telephone number could be a varchar or int). I
think that being able to specify access controls with respect to roles and
semantic data types is a powerful concept. It would go hand in glove with
the SQL-over-HTTP abstraction efforts under consideration.

I hope this makes sense.

Thanks,
Shaq


On 3/19/12 7:00 AM, "Francis Irving" <francis at flourish.org> wrote:

>Use cases in theory come up in ScraperWiki - in practice hard wiring
>API keys between specific scrapers and other data stores / APIs works
>well enough. 
>
>In reverse, private scrapers in ScraperWiki have an API key that gives
>access just to that store from other services.
>
>Very granular and lowest common denominator... But works now.
>
>Anything else will need to agree on a higher level structure. I've not
>seen enough real use cases to know how that should work.
>
>What's your use case?
>
>Francis
>
>On Sun, Mar 18, 2012 at 11:19:47PM -0400, Haq, Salman wrote:
>> Personally, I'm investigating role-based methods.
>> 
>> In my use-case, I can envision a vast number of 'agents' each associated
>> with some notion of 'identity' and 'role' trying to access data with the
>> policy enforcement mechanism working transparently. It's possible the
>> queries may complete fully, not at all, or partially, based on the
>> decision made by the policy enforcement mechanism. The important thing
>>is
>> that the decision has be made in 'real-time'.
>> 
>> XACML is a standard language for describing access policies. It affords
>>a
>> lot of flexibility but at the cost of verbosity and complexity. Are
>>there
>> other languages for this problem domain?
>> 
>> Also, do others have different use cases?
>> 
>> 
>> Thanks,
>> Shaq
>> 
>> 
>> On 3/18/12 9:00 PM, "Francis Irving" <francis at flourish.org> wrote:
>> 
>> >That's an excellent question!
>> >
>> >I'm mildly worried that every data hub, and indeed every piece of
>> >enterprise SaaS!, is inventing its own access control method.
>> >
>> >I've no idea what the best answer is. Anyone?
>> >
>> >Francis
>> >
>> >On Sun, Mar 18, 2012 at 08:38:33PM -0400, Haq, Salman wrote:
>> >> Hi all,
>> >> 
>> >> The Data Protocols group is engaged in very useful work. I was
>>curious
>> >>if there had been any discussions about access control mechanisms.
>> >>Inevitably, any database is usually tempered by access filters and if
>> >>people have any ideas about that, I would like to know. I skimmed the
>> >>archives about this topic but didn't find anything relevant.
>> >> 
>> >> Thanks,
>> >> Shaq
>> >> Architect, Neustar Inc.
>> >> 
>> >> 
>> >
>> >> _______________________________________________
>> >> data-protocols mailing list
>> >> data-protocols at lists.okfn.org
>> >> http://lists.okfn.org/mailman/listinfo/data-protocols
>> >
>> 
>> 
>> _______________________________________________
>> data-protocols mailing list
>> data-protocols at lists.okfn.org
>> http://lists.okfn.org/mailman/listinfo/data-protocols
>>