University of Glasgow
Threat modelling technique for GDPR compliance based on logical reasoning
Abstract
dc:description.abstractData-driven applications and services are increasingly being deployed across various sectors, where they collect, aggregate, and process vast amounts of personal data from diverse sources on centralized servers. Consequently, safeguarding the privacy and security of this data is crucial. Since May 2018, the EU/UK’s General Data Protection Regulation (GDPR) has necessitated sophisticated compliance models. Current threat modeling techniques, however, do not adequately address GDPR compliance, particularly in complex systems where personal data is collected, processed, manipulated, and shared with third parties. This thesis proposes a comprehensive solution to develop a threat modeling technique that addresses and mitigates non-compliance threats by integrating GDPR requirements with existing security and privacy modeling techniques, namely STRIDE (Spoofing, Tampering, Repudiation, Information Disclosure, Denial of Service, and Elevation of Privilege) and LINDDUN (Linking, Identifying, Non-repudiation, Detecting, Data Disclosure, Unawareness, and Non-compliance). The proposed technique in this thesis introduces a new data flow diagram aligned with GDPR principles, develops a knowledge base for non-compliance threats, and employs an inference engine to reason about these threats using the developed knowledge base. Additionally, this thesis presents a practical solution for modeling GDPR compliance using Defeasible Logic Programming (DeLP), enhancing the robustness and reasoning capabilities of compliance models in real-world scenarios. To address the challenges of undecided outputs in logical reasoning, this work incorporates explicit priorities for conflicting rules and suggests related knowledge for queries in an incomplete knowledge base. Furthermore, the technique includes a threat mitigation mechanism that identifies reasons for non-compliance threats and recommends actions to mitigate them. This approach is demonstrated through case studies on Telehealth Services and Fitbit (i.e., health tracking devices), focusing on addressing non-compliance threats and resolving UNDECIDED query results. Finally, the complexity of the defeasible reasoning mechanism is analyzed, and its performance is compared across different query outcomes, namely "YES/NO/UNDECIDED," based on vertical and horizontal complexities. The findings indicate that DeLP offers a flexible and dynamic framework suitable for implementing GDPR in real-world settings, making a significant contribution to the fields of legal reasoning and compliance modeling. Additionally, our findings show that the inference engine efficiently identifies non-compliance threats, handles UNDECIDED query results, and suggests appropriate threat mitigation measures.
Degree
thesis:*- Level dc:type.qualificationlevel
- PhD
- Grantor dc:publisher.institution
- University of Glasgow
- Year dc:date.issued
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Naila, Naila
Subjects
dc:subject × 1Rights
- Language dc:language
- en
Identifiers
dc:identifier.*- DOI dc:identifier.doi
- https://doi.org/10.5525/gla.thesis.84833