TruthTrust: Truth Inference-Based Trust Management Mechanism on a Crowdsourcing Platform

Area of Science:

• Computer Science
• Artificial Intelligence
• Crowdsourcing Security

Background:

• Crowdsourcing platforms face significant challenges from malicious workers submitting erroneous data to cheat or sabotage tasks.
• Existing defense mechanisms like verification, data analysis, and trust models have limitations in resource consumption, worker motivation, and detecting collusion attacks.

Purpose of the Study:

• To propose a new defense model, TruthTrust, to effectively defend against collusion attacks in crowdsourcing platforms.
• To ensure the authenticity of information and improve the overall integrity of crowdsourcing processes.

Main Methods:

• Developed a complete life cycle system for user interaction and worker recommendation.
• Defined separate trust values for workers and credence for requesters.
• Established a trust model based on the CRH framework to calculate truth values and weights.
• Implemented a reverse mechanism to enhance resistance against attacks.

Main Results:

• TruthTrust effectively defines global properties for workers and requesters.
• The proposed reverse mechanism improves resistance to malicious attacks.
• Extensive experiments demonstrate TruthTrust's significant outperformance over state-of-the-art approaches.

Conclusions:

• TruthTrust provides a robust solution for defending against collusion attacks in crowdsourcing.
• The model enhances the effective task completion rate and trustworthiness of crowdsourcing platforms.
• This approach contributes to a more reliable and efficient crowdsourcing ecosystem.