An Arbitrator Agent for EPrivacy

1
An Arbitrator Agent for E-Privacy

• Yingxin He
• Faculty of Computer Science,
• Dalhousie University

2
Outline

• Introduction
• Literature Survey
• Architecture of Arbitrator Agent
• Implementation of Arbitrator agent
• Negotiation Scenarios
• Conclusion
• Future Work

3
Introduction

• What is e-privacy?
• - The communication of personal data to others
  is under
• the control of the owner.
• What is P3P?
• - The Platform for Privacy Preferences Project
  (P3P).
• - P3P enables websites to express their privacy
  practices in
• a standard machine-readable format, XML
  format.

4
Introduction

• Motivation of this research
• - Todays e-commerce sites
• do not support negotiation.
• - Advantages of online
• negotiation.
• larger base of potential users
• time and cost savings

5
Literature Survey

• Online Privacy Agent (OPA)
• - It is the first implementation of P3P agent
  at IBM
• Almaden Research Center in 1999.
• A Sever-Centric P3P Implementation
• - It is an implementation of P3P agent at IBM
  
• Almaden Research Center in 2003.
• Client-Side Agent-Based E-Privacy Architecture
• - Proposed in 2004.

6
OPA

• Supports negotiation of personal information
• Good Start and Valuable foundation.
• Not suitable for private information
• negotiation.
• tree representation

7
OPA

• tree representation
• f1 (convertible, (drivermother),
  (seasonwinter))

8
Server-centric implementation
Client Side
Internet
Web Side
Server Side Logic
1. Send preference and URI of a web page
2. Preference and web page URI
Web server
Client Browser
3. SQL query
4. Query result
5. Send result of matching preference against
policy
6. Request web page if policy conforms to
preference
9
Service-centric implementation

• Comments on server-centric implementation
• - database technology speeds up the
• performance.
• - need a great amount of trust by the
• users.

10
Client-Side Agent-Based E-Privacy Architecture
External Privacy-related Information Feeds
Arbitrator Agent
Negotiation information
consultations
Law/ Governance issues/ reputations
Personal Context Agent
Regulatory Agent
Private data matching result
consultations
User Actions / Data warnings
Monitor Agent
P3P Agent
11
Architecture of Arbitrator Agent

• Negotiation Terminologies
• Ontology Retrieval and Tree Representation
• Negotiation Process
• Negotiation Strategies
• Comparison System

12
Negotiation Terminologies

• Purpose
• is a predefined value of P3P element
  ltPURPOSEgt, such as, current, admin, tailoring,
  contact, telemarketing,
• Recipient
• is a predefined value of P3P element
  ltRECIPIENTgt,
• such as, ours, delivery, same, public,

13
Negotiation Terminologies

• Retention
• value is a predefined value of P3P element
  ltRETENTIONgt, such as, no-retention,
  stated-purpose, business-practices,
• length is the actual retention length, for
  example, 2-week,
• 1-month,
• Usage
• A usage looks like
• U (current, develop, ours,
  (stated-purpose, 2-weeks))

14
Negotiation Terminologies

• Private Information
• is a private data that is defined
  in P3P, such as, user.name, user.age,
• Preference rule
• is a private data, is one of
  the usages under which d can be released.
• Sub-preference rule

15
Negotiation Terminologies

• Request
• d is a private data requested by a website,
  is one of the usages under which the
  requested data will be collected.
• Sub-request
• Counter-offer
• d is a private data, and is a usage.

16
Client-Side Agent-Based E-Privacy Architecture
Arbitrator Agent
Negotiation information
consultations
Law/ Governance issues/ reputations
Personal Context Agent
Private data matching result
P3P Agent
17
Conceptual OWL representation of P3P data
Hierarchy
ReleaseData
business
dynamic
thirdparty
user


dynamic. http
dynamic. cookie
business. name
business. department

thirdparty. name
thirdparty. bdate
user. gender

user. home-info
user. name
user. bdate


user. name. family
user. name. given
user. name. nickname
user. home-info. postal
user. home-info. telecom


user. home-info. postal. street
user. home-info. postal. city
user. home-info. postal. country
18
How to build preference trees?

• each data set together with its preference rules
  is represented as a tree.
• the leaves are usage sets.
• the root and intermediate nodes are private
• data items.
• a data is added to the preference tree if and
  only if it has a preference rule or at least one
  of its child elements has a preference rule.

19
How to build preference trees?
20
Advantages of tree representation

• Well organized data sets
• Easy to find a parent or a child of a
• data
• Suitable for negotiation

21
Negotiation Process
Arbitrator
Negotiator
Session ID, ask for connection
Session ID connection established
Compare requests to users preference rules
Session ID, accept requests, reject requests,
counter-offers
Evaluate counter-offers
Session ID accept counter-offers, reject
counter-offers, requests

• Negotiation progress
• Finite number of negotiation steps and rounds
• Conclude negotiation
• Maintain the knowledge repository

22
Negotiation Strategies

• Server-centric strategy
• - same data item
• - child elements
• - collaborate with the user
• Client-centric strategy
• - same data item
• - child elements
• - substitute data

23
Comparison System

• Data-distance function
• 0, if d d
• gt0, otherwise
• Purpose/Recipient-distance function
• 0, if
• gt0, otherwise

24
Comparison System

• Retention-distance function
• 0, if and only if
  
• gt0, otherwise
• Total-distance function

25
Implementation of Arbitrator Agent

• Programming language
• - JAVA
• Test Server

26
Negotiation Scenarios

• Scenario 1 Automatic negotiation
• Scenario 2 Collaborate with the user
• Server policy
• We collect user.login.id and user.login.password
  when you administrate on our site, so that you
  can access your information.
• We collect user.bdate and user.gender to tailor
  our site, so that we can highlight products
  related to your interests.

27
Scenario 1 Automatic negotiation
Negotiator
Arbitrator
28
Scenario 2 Collaborate with the user
Negotiator
Arbitrator
29
Conclusion

• The preference tree structure is the key
  improvement over the tree representation of OPA.
• The comparison system and negotiation strategies
  of current implementation is very promising

30
Future Work

• Electronic Contract
• Security
• Combination with supply chain negotiation
• Usability and delay evaluation

31
Acknowledgment

• Supervisors Dr. Jutla and Dr. Bodorik
• Committee members Dr. Riordan and
• Dr. McAllister
• Coworkers Mr. Yanjun Zhang, Mr. Deyun Gao, Ms.
  Liming Xu, and Mr. Xuehai Wang
• Graduate Scholarship of Saint Marys
  University
• My Family

32
Thank you!
33
Negotiation Scenario

• U1 (admin, develop, ours, (stated-purpose,
  1-year))
• U2 (pseudo-decision, tailoring, ours,
  (stated-purpose, 1-year))
• U1 (admin, develop, ours, (stated-purpose,
  1-year))
• U2 (pseudo-decision, tailoring,
  pseudo-analysis, develop,
• ours, (indefinitely, 8))
• U3 (pseudo-analysis, pseudo-decision,
  develop, ours,
• (stated-purpose, 1-year))

Negotiator
Arbitrator
34
Negotiation Scenario
Negotiator
Arbitrator
35
Negotiation Scenario
Negotiator
Arbitrator