Robust and reusable LINDDUN privacy threat knowledge

Abstract

Privacy threat modeling is an intrinsically complex analysis task that requires expertise in sophisticated privacy threats, their harms and implications, as well as potential mitigations. To support both novices and experts in attaining a desired degree of rigor and completeness in their analysis, supporting materials such as privacy threat trees and threat examples are crucial as they consolidate and harmonize the complete spectrum of threat characteristics, and as such assist with the broader uptake of privacy threat modeling practices.

However, the existing knowledge structures, taxonomies, and trees used in privacy threat analysis prove to have limited use in practice. They are either too broad and generic, or too tightly coupled to a specific modeling approach (dfds) or to a specific threat elicitation method (e.g., per-element). In addition, current privacy threat knowledge structures suffer from semantic ambiguity. Finally, existing structures are too rigid to support evolution, thus hindering the incorporation of emerging privacy threats.

This article introduces three contributions to address these shortcomings: (i) it defines the metamodel to express threat knowledge in the form of threat types, elicitation criteria, examples, and additional metadata; (ii) it discusses its application to the privacy threat knowledge of the linddun privacy threat modeling framework; and (iii) it introduces the automated knowledge management tools comprised of extraction logic that allows more flexible adoption in different privacy analysis approaches, and that fundamentally supports continuous evolution and refinement of this privacy threat knowledge. A major outcome is the updated linddun privacy threat knowledge which completely subsumes earlier versions and provides more rooted support for adoption, refinement, and continuous evolution.