Privacy-preserving crowd-sensed trust aggregation in the user-centeric internet of people networks

Abstract

Today we are relying on the Internet technologies for various types of services ranging from personal communication to the entertainment. The online social networks (Facebook, twitter, youtube) has seen an increase in subscribers in recent years developing a social network among people termed as the Internet of People. In such a network, subscribers use the content disseminated by other subscribers. The malicious users can also utilize such platforms for spreading the malicious and fake content that would bring catastrophic consequences to a social network if not identified on time. People crowd-sensing on the Internet of people system has seen a prospective solution for the large scale data collection by leveraging the feedback collections from the people of the internet that would not only help in identifying malicious subscribers of the network but would also help in defining better services. However, the human involvement in crowd-sensing would have challenges of privacy-preservation, intentional spread of false high score about certain user/content undermining the services, and assigning different trust scores to the peoples of the network without disclosing their trust weights. Therefore, having a privacy-preserving system for computing trust of people and their content in the network would play a crucial role in collecting high-quality data from the people. In this paper, a novel trust model is proposed for evaluating the trust of the people in the social network without compromising the privacy of the participating people. The proposed systems have inherent properties of the trust weight assignment to a different class of user i.e. it can assign different weights to different users of the network, has decentralized setup, and ensures privacy properties under the malicious and honest but curious adversarial model. We evaluated the performance of the system by developing a prototype and applying it to different online social network dataset.