Description

Security has emerged as one of the most critical quality attributes in modern software systems. To promote security by design, as part of the SCAM research project a recommender system is under development to assist in the secure design of software systems. Evaluating the effectiveness and reliability of the recommender system necessitates detailed evaluation scenarios. However, these scenarios typically demand domain-specific expertise, which is costly and challenging to gather.

This thesis will develop a methodologically sound approach for systematically utilizing ChatGPT-4 in the generation of evaluation scenarios. The approach will be tailored to ensure scientific rigor, reproducibility, and relevance to the domain of secure software engineering.

Expected Contributions:

• Novel Methodology: A structured approach for employing ChatGPT-4 as a tool for generating scientifically valid evaluation scenarios in software engineering, particularly for secure system design.
• Concept Evaluation: The first empirical evaluation of the recommender system for secure system design using ChatGPT-generated scenarios.
• Data Refinement: Enhanced and refined datasets for both future recommender system training and further research in the domain of secure software design.

This thesis is only available as a master thesis.