eyeprolog

ODRL policy

May this dataset be used for research?

An ODRL policy arrives as an RDF graph: one permission, one target, one action, and one purpose constraint. EyeProlog reads the graph as ordinary rdf/4 facts, decides whether the permission applies, and writes the decision back as RDF.

Run it in the playground · Source

The question

ODRL, the W3C Open Digital Rights Language, describes what may be done with an asset: who may perform which action, on what target, under which constraints.

Here the question is narrow and concrete: does the policy permit use of the dataset when the purpose is research? The answer must come from the policy graph itself, not from code that already knows it.

The policy as facts

rdf(iri('https://example.org/policy'), iri('http://www.w3.org/ns/odrl/2/permission'), iri('https://example.org/permission'), default_graph).
rdf(iri('https://example.org/permission'), iri('http://www.w3.org/ns/odrl/2/target'), iri('https://example.org/dataset'), default_graph).
rdf(iri('https://example.org/permission'), iri('http://www.w3.org/ns/odrl/2/action'), iri('http://www.w3.org/ns/odrl/2/use'), default_graph).
rdf(iri('https://example.org/permission'), iri('http://www.w3.org/ns/odrl/2/constraint'), iri('https://example.org/purpose-constraint'), default_graph).
rdf(iri('https://example.org/purpose-constraint'), iri('http://www.w3.org/ns/odrl/2/leftOperand'), iri('http://www.w3.org/ns/odrl/2/purpose'), default_graph).
rdf(iri('https://example.org/purpose-constraint'), iri('http://www.w3.org/ns/odrl/2/operator'), iri('http://www.w3.org/ns/odrl/2/eq'), default_graph).
rdf(iri('https://example.org/purpose-constraint'), iri('http://www.w3.org/ns/odrl/2/rightOperand'), iri('https://example.org/research'), default_graph).

Each RDF triple becomes one rdf(Subject, Predicate, Object, Graph) fact. The constraint reads: purpose equals research.

The decision rule

odrl_policy_decision(permit(use, research, dataset)) :-
  rdf(
    iri('https://example.org/policy'),
    iri('http://www.w3.org/ns/odrl/2/permission'),
    Permission,
    default_graph
  ),
  rdf(
    Permission,
    iri('http://www.w3.org/ns/odrl/2/target'),
    iri('https://example.org/dataset'),
    default_graph
  ),

The rule continues with the action and the three parts of the constraint. The variables Permission and Constraint join the triples: every condition must hold for the same permission node, not for unrelated ones.

The answer

eyeprolog examples/odrl-policy.pl

result_rdf(iri('https://example.org/result'), iri('https://example.org/decision'), iri('https://example.org/permit'), default_graph).
result_rdf(iri('https://example.org/result'), iri('https://example.org/action'), iri('http://www.w3.org/ns/odrl/2/use'), default_graph).
result_rdf(iri('https://example.org/result'), iri('https://example.org/purpose'), iri('https://example.org/research'), default_graph).
result_rdf(iri('https://example.org/result'), iri('https://example.org/target'), iri('https://example.org/dataset'), default_graph).

The decision is permit, written as four RDF triples that a graph store can load directly.

The reason

step(odrl_policy_decision(permit(use, research, dataset)),
     rule(8),
     ['Permission' = iri('https://example.org/permission'),
      'Constraint' = iri('https://example.org/purpose-constraint')],
     [rdf(iri('https://example.org/policy'), iri('http://www.w3.org/ns/odrl/2/permission'), iri('https://example.org/permission'), default_graph),
      ...]).

The permit holds by the decision rule (clause 8) with Permission bound to the permission node and Constraint to the purpose constraint. Its uses are exactly the seven policy triples, so an auditor sees which parts of the graph the decision rests on.

The check

condition('C1', resolution, ok, 12).
condition('C4', coverage, ok, 15).
condition('C7', relevance, ok, 16).
trusted(0).
claims(4).
verdict(checked).

All 12 steps replay against the program, the 4 claims are covered, and nothing is taken on trust.

Try it

Change the right operand from research to marketing and run again: the rule no longer fires, and there is no permit to report.
Add a second permission node for another dataset and check that the join on Permission keeps the two apart.
Compare with the richer ODRL policy reasoning deck, which adds prohibitions, duties, and conflicts.

Learn more in The Art of EyeProlog, Chapter 15.

This site is open source. Improve this page.