Privacy conflict analysis in web interaction models

Inglis, Peter (2022) Privacy conflict analysis in web interaction models. PhD thesis, University of Glasgow.

Full text available as:
[thumbnail of 2022inglisphd.pdf] PDF
Download (7MB)

Abstract

User privacy has become an important topic with strong implications for the manner by which software systems are designed and used. However, it is not a straightforward consideration on how the instrumentation of data processing activities contribute to the privacy risk of data subjects when interacting with data processors online. In this work, we present a series of methods to assist Data Protection Officers (DPOs) in the modelling and review of data processing activity between data processors online. We articulate an awareness formalism to model the knowledge gain of data processors and the privacy expectations of a data subject. Privacy conflict is defined in this work as an event where the expectations of the data subject do not align with the data processors knowledge gain resulting from data processing activity.

We introduce a Selenium workflow for the elicitation of data processing activity of web services online in the creation of an information flow network model. We further articulate a series of privacy anti-patterns to be matched as attributes on this model to identify data processing activity between two data processors facilitating conflict between data subjects and processors. Each anti-pattern illustrates a distinct manner by which conflict can arise on the information flow model. We define privacy risk as the ratio of third party data processors that facilitate an anti-pattern to the total number of third party data processors connected to a first party data processor. Risk in turn quantifies the privacy harm a data subject may incur when interacting with data processors online.

Pursuant to the reduction of privacy risk, we present a multi objective approach to model the inherit tensions of balancing the utility of a data subject against the cost incurred by a data processor in the removal of anti-patterns. We present our approach to first elicit the Pareto efficient set of anti-patterns, before operating on a utility function of programmable biases to output a single recommendation. We evaluate our approach against trivial selection strategies to reduce privacy risk and illustrate the key benefit of a granular approach to analysis. We conclude this work with an outlook on how the work can be expanded along with critical reflections.

Item Type: Thesis (PhD)
Qualification Level: Doctoral
Colleges/Schools: College of Science and Engineering > School of Computing Science
Supervisor's Name: Omoronyia, Dr. Inah
Date of Award: 2022
Depositing User: Theses Team
Unique ID: glathesis:2022-82875
Copyright: Copyright of this thesis is held by the author.
Date Deposited: 13 May 2022 15:19
Last Modified: 13 May 2022 15:20
Thesis DOI: 10.5525/gla.thesis.82875
URI: https://theses.gla.ac.uk/id/eprint/82875

Actions (login required)

View Item View Item

Downloads

Downloads per month over past year