Towards a policy language for humans and computers

1
Towards a policy language for humans and computers

• Vicky Weissman
• Joint work with Carl Lagoze

2
The big picture

• A policy says that under certain conditions an
  action, such as downloading a file, is permitted
  or forbidden.
• Digital content providers want
• to write policies about how their works may be
  accessed, and
• to have those policies enforced.

3
Diverse apps same need

• Because we cant regulate access to online
  content with precision
• Digital libraries cant put certain content
  online it might violate IP laws.
• The Greek Orthodox Archdiocese of America is wary
  of defamation.
• Cultural traditions arent respected. (Australian
  Aboriginal communities often restrict access to a
  clan or gender.)

4
XrML to the rescue

• XrML is a XML-based language for writing
  policies.
• The specification includes an algorithm that
  determines if a set of policies imply a
  permission.
• Idea write policies in XrML, enforce them using
  the algorithm.

5
Industry likes XrML

• XrML endorsed by Adobe, Hewlett-Packard,
  Microsoft, Xerox, Barnesandnoble.com, MPEG
  International Standards Committee
• Microsoft and others plan to make XrML-compliant
  products.
• Will tomorrows OS, DVD player, enforce XrML
  policies?

6
XrML Shortcomings

• Usability
• To read/write policies in XrML requires a
  significant amount of training.
• Even with training, writing policies is
  non-trivial and reading XrML policies is
  difficult.

7
A partial solution

• Build a nice interface.
• Users enter policies through the interface and
  then their input is translated to XrML.

UI
User Input
Translator to XrML
User Input in XrML
8
Problem

• There probably isnt a single interface that is
  appropriate for all users.
• So, we need an interface for each user community
  (e.g. musicians, publishers,)

UI1
UIn
User Input
Translator1 to XrML

Translatorn to XrML
User Input in XrML
9
Another problem

• Whose going to write the translations?
• Presumably, the UI designer.
• So, each UI writer is going to have to learn XrML
  and write a translator from input via their UI to
  XrML policies?
• There must be a better way!

10
Our solution

• Create a language that the UI designers can learn
  quickly and use easily.
• Then provide a translation from the intuitive
  language to XrML.

11
The big picture
Let R represent the new language.
UIn
UI1
.
User Input
Translatorn to R
Translator1 to R
.
User Input in R
Translator to XrML
12
Benefits of this approach

• If industry decides to enforce a new language,
  only one translation changes.

UIn
UI1
.
User Input
Translatorn to R
Translator1 to R

User Input in R
Translator to new language
13
Benefits of this approach

• Similarly, if some products use XrML, others
  ODRL, , need 1 new translator/language.

UIn
UI1
.
User Input
Translatorn to R
Translator1 to R

User Input in R
Translator to ODRL
Translator to XrML
14
Goal

• To create a language that is at least easier to
  use than XrML.
• Ideally, find a language that is easy to use.
• Big Idea Base the policy language on a natural
  language one that non-experts already know.
• We use English, but expect that our results
  readily translate to other (human) languages.

15
Rosetta

• We call the new language Rosetta.
• Rosetta is essentially a set of templates for
  creating English sentences.
• A sentence is in Rosetta, if we can create it by
  filling-in one of the templates with appropriate
  values.

16
Simple templates

• Rosetta includes the templates
• is .
• is .
• may .
• Given these templates, I claim that we can write
  the sentence Alice is smart.

17
Simple templates

• Rosetta includes the templates
• is .
• Alice is smart .
• is .
• may .
• Given these templates, I claim that we can write
  the sentence Alice is smart.

18
Simple templates

• Rosetta includes the templates
• is .
• is .
• may .
• Similarly, we can write Bob is a student.

19
Simple templates

• Rosetta includes the templates
• is .
• is .
• Bob is a student .
• may .
• Similarly, we can write Bob is a student.

20
Simple templates

• Rosetta includes the templates
• is .
• is .
• may .
• Its easy to see that
• Bob may watch Finding Nemo.
• is in Rosetta, because it matches the third
  template.

21
Simple templates

• Rosetta includes the templates
• is .
• is .
• may .
• Other simple templates in Rosetta can be used to
  capture sentences such as
• Every employee is trusted and
• Clark Kent is Superman.

22
Conditionals

• Rosetta includes if then statements of the
  form
• if and and then ss,
• where ss simple statement
• E.g. If today is Saturday and Alice is good, then
  Alice may watch Finding Nemo.

23
Empirical Observations

• Simple and if then sentences seem to be
  sufficiently expressive to capture most (all?)
  policies of practical interest.
• But capturing all simple and if then sentences
  is non-trivial.
• In particular, pronouns and prepositional phrases
  are difficult to support.

24
Pronouns

• Consider the statement if a toddler kicks Alice,
  then she is angry.
• Who is angry?
• Answer 1 the toddler, otherwise she wouldnt
  have kicked Alice.
• Answer 1 Alice, because she has been kicked.
• Bottom line Sentence is ambiguous, so we dont
  know how to translate it.

pronoun
25
Our solution Labels

• If Alice is angry, then we can write the sentence
  without using a pronoun.
• if a toddler kicks Alice, then Alice is angry.
• Otherwise, we replace the pronoun with a label
  that is associated with the toddler.
• if a toddler t kicks Alice, then t is angry
• Labels are not part of standard English, but seem
  fairly intuitive.

26
Prepositions

• Prepositions can also cause ambiguity.
• E.g. Alice designed the library in London.
• Sentence could mean that Alice designed the
  library that is in London or that Alice designed
  the library when she was in London.
• We choose the first interpretation, but this
  might not be the best solution.
• Maybe we should ask the writer?

Prepositional Phrase
27
Expressivity

• If we include propositions in Rosetta, then
  Rosetta is almost as expressive as XrML.
• Also, its easy to extend Rosetta to include all
  of XrML, but the sentences are a bit unwieldy.
• E.g. if the statements s1,, sn imply that Alice
  is good, then she may watch Finding Nemo.

28
Beyond XrML

• We can easily extend Rosetta to include sentences
  with negation.
• E.g. If Alice is not good, then she may not watch
  finding Nemo.
• Since XrML does not support negation, we couldnt
  translate the extended Rosetta to XrML.

29
Summary

• XrML is a policy language that is difficult to
  use, but will likely be enforced automatically by
  many next-generation products.
• Rosetta is a first-step towards an appropriate
  front end for XrML.
• It can serve as a front-end to other policy
  languages as well.
• Future work includes usability testing.