Inferring Trust Relationships in Web-Based Social Networks
Abstract
The growth of web-based social networking and the properties of those networks have created great potential for producing intelligent software that integrates a user's social network and preferences. This research focuses on the concept of trust in social networks. The goal of the work is to use explicit trust ratings that describe direct connections between people in social networks and compose this information to infer the trust that may exist between two people who are not directly connected. The paper presents two variations on an algorithm to make this calculation in networks in which users rate one another on a binary scale (trusted or not trusted). The authors begin by presenting a definition of trust and illustrating how it fits in with making trust ratings in web-based social networks. For both algorithms, the objective is to infer trust values that are accurate to the person for whom they are calculated. They introduce each algorithm in detail, followed by a theoretical analysis that shows why highly accurate results can be expected. This is reinforced through simulation that demonstrates the correctness in simulated networks. Finally, they demonstrate the potential of using inferred trust values to create trust-aware applications through a prototype of TrustMail, an e-mail client that uses trust ratings as a mechanism to filter e-mail.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 2006
- Accession Number
- ADA447930
Entities
People
- James Hendler
- Jennifer Golbeck
Organizations
- University of Maryland