Abstract

This thesis deals with reputation and trust systems in IoT-Feds marketplaces, and particularly with an efficient way to calculate and update the reputation score of the data sources, products, providers, and federations in the federated data marketplace developed by project IoTFeds.

First, a literature review was conducted on reputation and trust metrics and their applications in online marketplaces. Existing ways to compute the reputation scores in both centralized and decentralized manners were studied. Subsequently, the reputation and trust system for the IoTFeds marketplace is presented. The reputation of a data product is calculated as the aggregation of both objective and subjective scores in a context-specific and user-oriented way.

Then the framework of experimental assessment of this reputation mechanism is developed, and a series of experiments are carried out, by means of which the functionality of the reputation mechanism is verified. The experiments confirm that the reputation score of a product is fluctuating as expected after some transactions based on quality of the data product as well as on the objective and subjective scores submitted.

Finally, a reputation-correction approach was developed that aims to identify liar providers that over-rate the quality of their resources and submit falsely high objective scores to augment their trust. The experiments reveal that this corrective set handles successfully the discrepancies between objective and subjective reputation metrics whenever arising. Last, through references in the bibliography, we illustrate the advantages and disadvantages of the blockchain, and some of the most common attacks that could affect the blockchain and thus reveal vulnerabilities in the architecture considered.