Exploration algorithms for discoverable and undiscoverable decentralised online social networks
Ara, Safina Showkat
Ara, Safina Showkat
Loading...
Publication Date
2021-12-09
Type
Thesis
Downloads
Citation
Abstract
Decentralised online social networks (DOSNs) are a solution to privacy and security problems encountered in online social networks. In DOSNs, a user controls their data and chooses personal storage for their social network data. Concerning the concept of “discoverability”, most online social networks are discoverable networks, as the features of and entities within these networks are recognisable or predictable. In the absence of a centralised entity, information retrieval becomes difficult in discoverable DOSNs, and even more so in undiscoverable ones. The complexity of such a problem arises from the limited access of a user to the social network data: even in a discoverable DOSN, there is no centralised entity with access to all the social network data. Crowdsourcing can help us to retrieve information in both discoverable and undiscoverable DOSNs. However, verification of the quality of workers is a major problem in crowdsourcing. Unfortunately, it is not easy to do this without hiring more workers. One can face several problems when exploring groups of users to find workers in DOSNs. Detecting experts and managing their evolution can be a difficult task, especially in highly complex environments. However, none of the existing solutions have completely focused on these issues in relation to a decentralised platform. In this thesis, we analyse a popular information retrieval problem known as expert search in a social network. Our goal is to present an algorithm for such a crowdsourcing-based search process which includes a solution for the worker selection problem, the task selection problem, and the reward distribution problem. Using experimental evaluation, we show that the proposed search algorithms can be as efficient as a greedy search algorithm with access to full social network information. After completion of an expert search, we aim to judge the performance of the workers based on their history of service, which as mentioned previously is not easy to do so without hiring other workers. We propose an Ant Colony Optimisation (ACO)-based reputation management system that can differentiate between good and bad workers. Using experimental evaluation, we show that the algorithm works well on the given scenario and efficiently differentiates workers with higher reputations. Moreover, we present an innovative model of crowdsourcing for complex tasks in order to mitigate the above problems. We replace the centralised coordinator by a blockchain and automate the decision-making process of the coordinator. We show that the proposed solution is secure and efficient, and that the computational overhead, that is due to employing a blockchain, is low.
Publisher
NUI Galway