Data Companion Set - Master's Thesis - Gabriel Gehrig

This companion set contains all research artifacts produced in support of the master's thesis: "LLM-based Privacy Specification and Data Flow Diagram Derivation for Privacy Analysis".

It accompanies two empirical studies evaluating LLM-based pipelines for automated derivation of:

(1) Data Flow Diagrams (DFDs) from functional requirements and

(2) privacy specifications (labels and neverFlows constraints) from legal requirements.

Contents:

Pipeline code - prototype implementation of the DFD derivation pipeline (RQ1) and privacy specification derivation pipeline (RQ2); provided as-is for reproducibility

Prompt templates - all phase-specific prompt templates used in the DFD derivation pipeline (RQ1) and the privacy specification pipeline (RQ2), including few-shot examples and output format constraints

Running Example - complete pipeline outputs, including intermediate phase outputs for the Travel Planner running example used throughout the thesis

User study materials - all survey pages as presented to participants, PHP and HTML implementation for SosciSurvey, raw study data, and analysis scripts

Validation schemas - Pydantic output validation schemas for each pipeline phase in both RQ1 and RQ2