eyeprolog

Ranked consumer-policy risks

Which clauses of this agreement put the consumer at risk, and which matter most?

A consumer agreement is written in ODRL, and the consumer’s needs are written in a DPV-style vocabulary, both as RDF. EyeProlog finds each clause that conflicts with a need, scores the risk, and returns one deterministic ranking with a mitigation for every risk.

Run it in the playground · Source

The question

ODRL describes what a service may do (permissions, duties, prohibitions). DPV, the W3C Data Privacy Vocabulary, describes privacy concepts such as risks and the needs of a data subject.

Put side by side, they let a program ask: does the agreement allow account removal without notice, terms changes with too little notice, sharing without explicit consent, or forbid data export? And if several risks apply, which should be fixed first?

Risk triggers are rules over the policy

risk_triggered(deletion_risk, Rule, _Need) :-
  rdf_link(ex('consumer-policy'), odrl(permission), Rule),
  rdf_link(Rule, odrl(action), ex('remove-account')),
  \+ duty_action(Rule, ex(inform)),
  \+ constraint_number(Rule, ex(noticeDays), _Days).

risk_triggered(terms_risk, Rule, Need) :-
  rdf_link(ex('consumer-policy'), odrl(permission), Rule),
  rdf_link(Rule, odrl(action), ex('change-terms-action')),
  duty_action(Rule, ex(inform)),
  constraint_number(Rule, ex(noticeDays), Days),
  rdf_number(Need, ex(minimumNoticeDays), Minimum),
  Days < Minimum.

A deletion risk is a removal permission with no duty to inform and no notice period. A terms risk is a terms change whose notice period is shorter than the consumer’s minimum.

Scoring and ranking

risk_report(Risk, Score, Level, Clause, Mitigation) :-
  risk_profile(_Profile, Risk, Rule, Need, Base, Mitigation),
  risk_triggered(Risk, Rule, Need),
  rdf_text(Rule, ex(clause), ClauseText),
  Clause = ClauseText,
  rdf_number(Need, ex(importance), Importance),
  Raw is Base + Importance,
  (Raw > 100 -> Score = 100 ; Score = Raw),
  risk_level(Score, Level).

risk_level(Score, high) :- Score > 79.
risk_level(Score, moderate) :- Score > 49, Score < 80.

The score is a base score plus the importance of the need, capped at 100. consumer_risk_report/1 collects every report with findall/3, sorts by inverted score and clause, and numbers the result, so ties always break the same way.

The answer

eyeprolog examples/odrl-dpv-risk-ranked.pl

result_rdf(iri('https://example.org/deletion-risk'), iri('https://example.org/rank'), literal('1', datatype('http://www.w3.org/2001/XMLSchema#integer')), default_graph).
result_rdf(iri('https://example.org/sharing-risk'), iri('https://example.org/rank'), literal('2', datatype('http://www.w3.org/2001/XMLSchema#integer')), default_graph).
result_rdf(iri('https://example.org/terms-risk'), iri('https://example.org/rank'), literal('3', datatype('http://www.w3.org/2001/XMLSchema#integer')), default_graph).
result_rdf(iri('https://example.org/portability-risk'), iri('https://example.org/rank'), literal('4', datatype('http://www.w3.org/2001/XMLSchema#integer')), default_graph).

The full output has 20 RDF triples: rank, score, level, clause, and mitigation for each of the four risks. Deletion scores 100 (high), sharing 97 (high), terms 85 (high), and portability 70 (moderate).

The reason

step(2 is 1 + 1, builtin, [], []).

The proof follows the ranking down to its arithmetic: the step above is one of the rank increments, recomputed by the checker. The collection of risk reports is different. It is recorded as one collected step for the whole findall/3, because a list of all solutions is a claim that nothing was missed, and that cannot be replayed step by step.

The check

condition('C1', resolution, ok, 72).
condition('C5', re_decision, ok, 31).
condition('C6', boundary_consistency, ok, 0).
trusted(1).
claims(20).
verdict(checked_with_obligations).

72 steps replay against the program and 31 built-in steps are recomputed. The verdict is checked_with_obligations: the proof is valid given one trusted boundary, the completed findall/3 that lists the four risk reports. The check report names it as an obligation/3 fact, so a reviewer knows exactly which claim to audit separately.

Try it

Add an inform duty and a notice period to the removal permission. Which risk disappears, and how do the remaining ranks shift?
Raise the importance of the portability need. Does its level change from moderate to high?
Compare with the ODRL policy reasoning deck, which decides individual requests instead of ranking risks.

Learn more in The Art of EyeProlog, Chapter 8.

This site is open source. Improve this page.