A trust model for advisor networks in multi-agent environments / Elham Majd
Multi-agent systems can break interactions in distributed and heterogeneous environments. One of the fundamental challenges in such settings is that agents can enter and leave the system at will; hence malicious agents may take advantage of others by behaving in an untrustworthy way. In this case, i...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
2015
|
| Subjects: | |
| Online Access: | http://studentsrepo.um.edu.my/5874/1/Elham_Majd%2D_Dissertation_(2015).pdf http://studentsrepo.um.edu.my/5874/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Multi-agent systems can break interactions in distributed and heterogeneous environments. One of the fundamental challenges in such settings is that agents can enter and leave the system at will; hence malicious agents may take advantage of others by behaving in an untrustworthy way. In this case, if an agent wants to interact with unknown provider agents, they need to request other agents to advise a trustworthy provider. The crucial issues are then how to rely on the information provided by advisor agents. A trust mechanism was proposed that measures and analyzes the trust value of advisors. In fact, the proposed mechanism measures the belief and disbelief value of each advisor in multi-agent environments utilizing reliability/ unreliability, reputation/disrepute of each interaction. In this mechanism, the aim was to select the trustworthy
provider agent through an advice of benevolent advisors in which the actions of advisors are accurately under analysis. The theoretical analysis was done in two parts;
first the validation of model was investigated by analyzing the average accuracy of model in calculating the trust and trust transitivity value among advisors and by comparison with other alternative models. Second, the average accuracy of our model in decision-making process was investigated by trust network game. The results denote that our approach outperforms current models in providing accurate credibility
measurements and computing an accurate trust mechanism for advisor agents in an advisor network, also presenting an accurate decision making process to choose the trustworthy provider. The experimental results also show the superior performance of our proposed model in comparison with other trust models. Applying this trust model can ensure critical transactions are performed more securely, such as those related to banking or e-commerce. |
|---|