Elimination of Subjectivity from Trust Recommendation

The opinion of a user about the trustworthiness of an entity is subjective to the disposition of trust of the user. This subjectivity may result in a distorted interpretation of the trustworthiness by other users. The objective of this project is to propose solutions for the elimination of subjectivity from trust evaluations.

Problem

A variety of trust-based systems have been developed that enable users to determine if the party they wish to transact with is trustworthy. Trust recommendation is a key technique that is utilized in trust-based systems for a user to determine the trustworthiness of an unknown party. A trust recommendation is an attestation of the trustworthiness of a user Carol by Alice to Bob, where Bob is a user who is not acquainted with Carol but maintains a trust relationship with Alice.

We present the argument that trust evaluation by each individual is subjective and thus when two individuals exchange a trust value, its meaning is distorted due to differences in their perception. For example, Alice may have suggested to Bob that the trustworthiness of Carol is 0.8 on the interval [0,1], which according to her subjective opinion may have been average trustworthiness. However, it is possible that Bob has a different perspective on trust values and regards 0.8 as a very high value. Thus subjectivity prevents the true meaning of Alice's recommendation from being conveyed to Bob.

Contributions

We have presented a method for the elimination of subjectivity from trust recommendation that takes advantage of trust scores given as percentiles, which are equally meaningful among two users. Experiments conducted on trust graphs have demonstrated that the method is effective for the elimination of subjectivity from trust recommendation. The method is non-intrusive and does not require any change in how users locally evaluate other users. Furthermore, the method does not involve any third party mediation, thus making it suitable for decentralized networks.

Contributors

Internal

• Omar Hasan
  
• Lionel Brunie
  

External

• Elisa Bertino
  
• Jean-Marc Pierson

Selected Publications

• Elimination of Subjectivity from Trust Recommendation. Omar Hasan, Lionel Brunie, Jean-Marc Pierson, and Elisa Bertino. The 3rd IFIP International Conference on Trust Management (TM 2009), West Lafayette, IN, USA. June 15 - 19, 2009. [PDF]

• Elimination de la Subjectivité dans la Recommandation de Confiance. Omar Hasan, Lionel Brunie, Jean-Marc Pierson, and Elisa Bertino. 5èmes Journées Francophones Mobilité et Ubiquité (UbiMob 2009), Lille, France. July 7 - 8, 2009.