Use this URL to cite or link to this record in EThOS:
Title: Recommending access control decisions to social media users
Author: Misra, Gaurav
ISNI:       0000 0004 6424 635X
Awarding Body: Lancaster University
Current Institution: Lancaster University
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Access from Institution:
Social media has become an integral part of the Internet and has revolutionized interpersonal communication. The lines of separation between content creators and content consumers have blurred as normal users have platforms such as social media sites, blogs and microblogs at their disposal on which they can create and consume content as well as have the opportunity to interact with other users. This change has also led to several well documented privacy problems for the users. The privacy problems faced by social media users can be categorized into institutional privacy (related to the social network provider) and social privacy (related to the interpersonal communication between social media users) problems. The work presented in this thesis focuses on the social privacy issues that affect users on social media due to their interactions with members in their network who may represent various facets of their lives (such as work, family, school, etc.). In such a scenario, it is imperative for them to be able to appropriately control access to their information such that it reaches the appropriate audience. For example, a person may not want to share the same piece of information with their boss at work and their family members. These boundaries are defined by the nature of relationships people share with each other and are enforced by controlling access during communication. In real life, people are accustomed to do this but it becomes a greater challenge while interacting online. The primary contribution of the work presented in this thesis is to design an access control recommendation mechanism for social media users which would ease the burden on the user while sharing information with their contacts on the social network. The recommendation mechanism presented in this thesis, REACT (REcommending Access Control decisions To social media users), leverages information defining interpersonal relationships between social media users in conjunction with information about the content in order to appropriately represent the context of information disclosure. Prior research has pointed towards ways in which to employ information residing in the social network to represent social relationships between individuals. REACT relies on extensive empirical evaluation of such information in order to identify the most suitable types of information which can be used to predict access control decisions made by social media users. In particular, the work in this thesis advances the state of art in the following ways: (i) An empirical study to identify the most appropriate network based community detection algorithm to represent the type of interpersonal relationships in the resulting access control recommendation mechanism. This empirical study examines a goodness of fit of the communities produced by 8 popular network based community detection algorithms with the access control decisions made by social media users. (ii) Systematic feature engineering to derive the most appropriate profile attribute to represent the strength or closeness between social media users. The relationship strength is an essential indicator of access control preferences and the endeavor is to identify the minimal subset of attributes which can accurately represent this in the resulting access control recommendation mechanism. (iii) The suitable representation of interpersonal relationships in conjunction with information about the content that result in the design of an access control recommendation mechanism, REACT, which considers the overall context of information disclosure and is shown to produce highly accurate recommendations.
Supervisor: Such, Jose M. Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral