Conformance to legal requirements

1. Conformance to legal requirements The last frontier for privacy research Luigi Logrippo Université du Québec en Outaouais University of Ottawa Keynote Talk, PST2012

2. Towards a process for producing software fromlegalrequirements

3. The last frontier … • Last for two reasons: • In the end, privacy IT systems must satisfy the law • This is a difficult goal • Because of the need to bridge the long distance between legal language and IT language and implementations A bridge? A bridge?

4. Why IT isn’t very good at building bridges • Humankind has been building bridges for millions of years • But IT is fairly new at this … Vestiges of a bridge to Sri Lanka (30Km) Reputed to be 1,700,000 yrs old …

5. The islands and the bridges legal compliance Enterprise Privacy Policy Privacy Law Legal area extract extract Legal requirements for privacy software Enterprise requirements for privacy software logical compliance Software area implement & verify Enterprise privacy software validate

6. Legal processes • Formulating laws and policies • Normative text • Establishing legal compliance of enterprise privacy policy to the law: • These processes are entirely in the legal domain, for lawyers

7. How many research areas remain for us? • At least five, let’s see     

8. Manual activities   • At the interface of the legal world and the IT world • Extract from law • Determine what part of the law is relevant for IT implementation • Express that part of the law in IT terminology • Extract from enterprise policy • Determine what part of the enterprise privacy policy is relevant for IT implementation • Express that part of the policy in IT terminology • Determine compliance of  to 

9. How are the manual activities done? • Since they are interface activities, they will require mixed teams of law experts and IT experts • It’s usual for IT people to work in this manner, at the requirement extraction phase

10. Current practice • Legal requirements are expanded into many detailed requirements • Legal offices are used to check that these detailed requirements represent a legally defendable implementation of the law • Long checklists result from this process, and the enterprises subject to the law will use the checklists rather than the original law • Checklists usually include all sorts of things, not only software requirements

11. Elements for requirement extraction • As always in computing, we have data structures and processes • Examples of relevant data structures: • Enterprise organization diagrams • Concept ontologies • Examples of processes: • Business processes • These exist in law, as they exist in enterprise policies, • They may be difficult to find • They will be more generic in law

12. Can we put order in something like this? • Royal Bank of Canada privacy policies: • “We use your personal and financial information for the purposes communicated to you in your agreement(s) with us, for example to: Verify your identity; Provide you with the financial products and services requested; Communicate to you any benefit, feature and other information about products and services you have with us; Respond to any special needs or inquiries you may have; Better understand your financial situation and determine your eligibility for products and services we offer; Manage our risks and operations; Meet regulatory and legal requirements; If we have your social insurance number or social security number, we may use it for tax related purposes if you hold a product generating income and share it with the appropriate government agencies. We may also share it with credit reporting agencies as an aid to identify you.”

13. Identifying the dependencies(Ghazinour and Barker, PAIS 2011) Royal Bank of Canada privacy policies: “We use your personal and financial information for the purposes communicated to you in your agreement(s) with us, for example to: Verify your identity; Provide you with the financial products and services requested; Communicate to you any benefit, feature and other information about products and services you have with us; Respond to any special needs or inquiries you may have; Better understand your financial situation and determine your eligibility for products and services we offer; Manage our risks and operations; Meet regulatory and legal requirements; If we have your social insurance number or social security number, we may use it for tax related purposes if you hold a product generating income and share it with the appropriate government agencies. We may also share it with credit reporting agencies as an aid to identify you.” Purpose ontology lattice for RBC privacy policies

14. Why a lattice(Ghazinour and Barker, PAIS 2011) This lattice arranges the purposes in an implication order. E.g. if one allows RBC to use personal information for mail distribution then one has also allowed them to use it for communication, marketing, and identity verification (more specific purposes)

15. Organization structure and scenarios in the law • Sarbanes Oxley - Section.2 : Audit (3) AUDIT COMMITTEE. • The term ‘‘audit committee’’ means a committee established by and amongst the board of directors of an issuer for the purpose of overseeing the accounting and financial reporting processes of the issuer and audits of the financial statements of the issuer, …. Issuer: company subject to SOX

16. IT interpretation • Sarbanes Oxley - Section.2 : Audit (3) AUDIT COMMITTEE. • The term ‘‘audit committee’’ means a committee established by and amongst the board of directors of an issuer for the purpose of overseeing the accounting and financial reporting processes of the issuer and audits of the financial statements of the issuer, …. Exercise: draw the class diagram

17. Processes • In law, they are usually defined only in terms of what they should achieve – Examples from SOX • (a) pertain to the maintenance of records that in reasonable detail accurately and fairly reflectthe transactions and dispositions … • (b) provide reasonable assurance that transactions are recorded as necessary to permit preparation of financial statements … • (c) provide reasonable assurance regarding prevention or timely detection of unauthorized acquisition, use or disposition … • Details are found in standards, professional and ‘best practices’ manuals

18. Generic extraction model for enterprise governance(Hassan and Logrippo, RELAW2009) Note specific purpose! Concepts found in normative text are to be mapped into these classes

19. Scanning normative text • The extraction process can be to carefully scan the law, standards, ‘best practices’, enterprise regulations, looking for elements that can be implemented in software • Concepts found should be mapped on an extraction model that can be the basis for software implementation • Conceptual graphs, lattices, UML

20. Joint work • The resulting formalized representations are interpretations of the original text • For IT specialists, the acceptance criterion is: • can this interpretation be implemented in software? • For Law specialists, the criterion is: • can this interpretation be defended in court?

21. We are interested in the intersection • Identify and formalize the intersection • Expand it as much as possible

22. Compliance of enterprise requirements to legal requirements legal compliance Enterprise Privacy Policy Privacy Law Legal area extract extract Legal requirements for privacy software logical compliance Enterprise requirements for privacy software  Software area implement & verify Enterprise privacy software validate

23. Compliance of enterprise requirements to legal requirements • What was a legal compliance process in the legal area becomes a logical compliance check in the software area • This can be performed by using model checkers of various kinds 

24. Proposal: A Logic-Based Process(Hassan and Logrippo) OK or counterexamples

25. Checking requirements on organization structure An organization with two main departments, incl. several processes Contains (Loans, PublishApplication) Contains (Loans, ReceiveFilledApp) Contains (Loans, Wapplication) Contains (Loans, JReceiveFilledApp) Contains (Loans, ConsentClient) Contains (Loans, LegalReasonException) Contains (Loans,ThankClient) Contains (Loans, DisposeData) Contains(OrderMgt, ReadApplication) Contains(OrderMgt, ValidateInfo) Contains(OrderMgt,SaveInfo) Formally defined Enterprise structure Legal Requirement The organization must include a process to dispose of data Model checker: yes, it is included

26. Checking requirements on process structure Next (ValidateInfo,SaveInfo) Next(ReadApplication, ValidateInfo) Next(Wapplication, JreceivedApp) Next(JReceivedApp,ConsentClient) Next(JReceivedApp,LegalReasonException) Next(ThankClient,DisposeData) Next(PublishApplication, ReceiveFilledApp) Next(ReceiveFilledApp,Wapplication) Next(ValidateInfo,WApplication) Next(WApplication,ReadApplication) Formally defined structure Legal requirement: Information received must later be disposed Model checker: following slide

27. Process non-compliance A path is found where information rec’d is saved

28. Implement &verify legal compliance Enterprise Privacy Policy Privacy Law Legal area extract extract Legal requirements for privacy software Enterprise requirements for privacy software logical compliance Software area  implement & verify Enterprise privacy software validate

29. Implement &verify • We are here in a familiar territory: • We have compliant software requirements and we must implement them and verify the implementation • Use existing software methods • But: are the enterprise requirements that were obtained so far sufficient to derive an implementation? • A lot of practical domain knowledge may still be necessary • Probably it cannot be assumed that inexperienced software developers can do this 

30. Maturity of SE methods • Unfortunately, the study of techniques to go from requirements to implementations is fairly recent and so not very mature IMOMO • Requirements engineering • We have been doing this for only about half a century …

31. Generic SE development method 31 • Major errors can be injected at every step • especially between requirements and behavior specs • Legal knowledge is probably still needed between steps Requirements (in natural or logic language) Specification of behavior Specification of implementation 31 Implementation

32. Validate implementation legal compliance Enterprise Privacy Policy Privacy Law Legal area extract extract Legal requirements for privacy software logical compliance Enterprise requirements for privacy software Software area implement & verify  Enterprise privacy software validate

33. Validate implementation • Is the resulting enterprise software compliant with the law? • This must be checked since errors can be injected in the implementation process • Existing software methods can be used to validate the implementation wrt legal requirements, perhaps the most practical is testing • Final testing is part of every engineering process • But, exactly what should be tested and how? • The checklist mentioned are not constructed as software test suites 

34. Certification • The end result should be certified software • Certified to be conformant to the law • What should the certification process be? • Most probably, test suites derived from the mentioned checklists • Many software vendors produce software that is claimed to be compliant • But can hardly be certified

35. Privacy by Design • PbD is embedded into the design and architecture of IT systems and business practices • It is not bolted on as an add-on • Privacy becomes an essential component of the core functionality being delivered • Privacy is integral to the system, without diminishing functionality • (source: Information and Privacy Commissioner of Ontario, Canada) • How can be build PbD in the software process?

36. How can we move forward? • Of course, it would help if normative text to be implemented in software was written in a different style … • E.g. legal text leaves much unspecified • The complex ontologies on which its interpretation depends are rarely specified • The increasing dependence on IT systems will lead legislators to include more IT language and structure in their normative style • Thus facilitating the extraction process

37. More mind-expanding ideas • Formalizing Privacy agreements • P3P and extensions • Developments in legal theory and practice • Legal formalization necessary to expand e-Business • e-Contracts, internationally formalized • e-Judgments • Privacy violations to be proved automatically by using automatically obtained factual evidence • Amends to be determined automatically, on the basis of objective law

38. Many open areas of research • Formal semantics of normative languages • Methods to extract ontologies and processes from normative text (see RELAW workshop series) • Methods to validate the result of the extraction process • Such methods will be domain-specific • Software Engineering issues, instantiated to the legal domain • Methods for validating compliance of an implementation to legal requirements • Leading to certification • The PbD software process

39. Access control and normative concepts • In his 2012 paper, S. Barker states: “In future work, we intend to consider the type of more general authorization model requirementsthat Jones and Sergot have considered and their representation of rich forms of non-standard access control models within our framework” In other words, he intended to pursue a generalization of his access control theory to include normative concepts. Steve Barker: Logical Approaches to Authorization Policies. Logic Programs, Norms and Action 2012: 349-373

40. Conclusions • I have attempted to identify the main issues related to the problem of compliance to legal requirements for privacy • Classify the issues, by means of a proposed ‘reference model’ • Some preliminary solutions and research ideas were also presented, as possible starting points