Measures of Privacy Protection on Social Environments

Abstract

Nowadays, online social networks (OSNs) have become a mainstream cultural phenomenon for millions of Internet users. Social networks are an ideal environment for generating all kinds of social benefits for users. Users share experiences, keep in touch with their family, friends and acquaintances, and earn economic benefits from the power of their influence (which is translated into new job opportunities). However, the use of social networks and the action of sharing information imply the loss of the users' privacy. Recently, a great interest in protecting the privacy of users has emerged. This situation has been due to documented cases of regrets in users' actions, company scandals produced by misuse of personal information, and the biases introduced by privacy mechanisms. Social network providers have included improvements in their systems to reduce users' privacy risks; for example, restricting privacy policies by default, adding new privacy settings, and designing quick and easy shortcuts to configure user privacy settings. In the privacy researcher area, new advances are proposed to improve privacy mechanisms, most of them focused on automation, fine-grained systems, and the usage of features extracted from the user's profile information and interactions to recommend the best privacy policy for the user. Despite these advances, many studies have shown that users' concern for privacy does not match the decisions they ultimately make in social networks. This misalignment in the users' behavior might be due to the complexity of the privacy concept itself. This drawback causes users to disregard privacy risks, or perceive them as temporarily distant. Another cause of users' behavior misalignment might be due to the complexity of the privacy decision-making process. This is because users should consider all possible scenarios and the factors involved (e.g., the number of friends, the relationship type, the context of the information, etc.) to make an appropriate privacy decision. The main contributions of this thesis are the development of metrics to assess privacy risks, and the proposal of explainable privacy mechanisms (using the developed metrics) to assist and raise awareness among users during the privacy decision process. Based on the definition of the concept of privacy, the dimensions of information scope and information sensitivity have been considered in this thesis to assess privacy risks. For explainable privacy mechanisms, soft paternalism techniques and gamification elements that make use of the proposed metrics have been designed. These mechanisms have been integrated into the social network PESEDIA and evaluated in experiments with real users. PESEDIA is a social network developed in the framework of the Master's thesis of the Ph.D. student, this thesis, and the national projects in Social Educational Environments during Childhood and Adolescence (TIN2014-55206-R) and Intelligent Agents for Privacy Advice in Social Networks (TIN2017-89156-R). The findings confirm the validity of the proposed metrics for computing the users' scope and the sensitivity of social network publications. For the scope metric, the results also showed the possibility of estimating it through local and social centrality metrics for scenarios with limited information access. For the sensitivity metric, the results also remarked the users' misalignment for some information types and the consensus for a majority of them. The usage of these metrics as part of messages about potential consequences of privacy policy choices and information sharing actions to users showed positive effects on users' behavior regarding privacy. Furthermore, the findings of exploring the users' trade-off between costs and benefits during disclosure actions of personal information showed significant relationships with the usual social circles (family members, friends, coworkers, and unknown users) and their properties. This allowed designing...