Trust Management for the Semantic Web

1
Trust Management for the Semantic Web

• Paper Review by
• Boanerges Aleman-Meza
• CSCI 8350 Enterprise Integration
• Semantic Web Course
• Large Scale Distributed Information Systems
  (LSDIS) Lab

2
Essentials

• How can information be trusted?
• anyone can publish anything
• many users, many statements
• How to assign credibility to statements and/or
  users in an automatical and scalable manner

3
Trust

• Examples of trust
• epinions
• ebay
• Levels of trust
• Site (or user) CNN vs. NPR vs. Fox
• Statement Howard quit PresidentialRace

4
Web of Trust
i
5
Assumptions

• Each user specifies a small set of users she or
  he trusts
• Local neighbors help in determining trust of
  distant neighbors

6
Web of Trust
0.8
B
0.6
Statement i
A
C
1.0
0.48

• belief value 0 . . 1
• An statement is
• accurate
• credible
• relevant

• User trust value 0 . . 1
• An user is
• trustworthy
• credible
• shares common interest

7
Web of Trust
0.8
B
0.6
Statement i
A
C
1.0
0.48

• Scalability Goal
• compute trust of A based on local information

8
Critique

• Is a user trustworthy in everything he/she says?
• Dimensions of trust, dimensions of belief
• Aggregation maximum? Really?

9
Reference

• 1 Matthew Richardson, Rakesh Agrawal, and
  Pedro Domingos (2003). "Trust Management for the
  Semantic Web," Proceedings of the Second
  International Semantic Web Conference, Sanibel
  Island, FL, 2003