An LLM-based Approach to Recover Traceability Links between Security Requirements and Goal Models
Jameleddine Hassine
Abstract
The recovery of requirements traceability links between goal models and requirements is crucial for ensuring alignment between stakeholder objectives and system specifications. Large Language Models (LLMs) show potential to transform automated traceability significantly, addressing challenges such as accurately capturing diverse relationships between requirements artifacts, and ensuring scalability and efficiency in large-scale software projects. In this paper, we propose an LLM-based approach to generate security-related traceability links between requirements (expressed in natural language) and goals (described as part of GRL models). We employ a Zero-Shot (0S) approach utilizing GPT-3.5-turbo, enhanced by employing a meticulously crafted prompt. The approach is implemented in a prototype tool, tailored for the textual GRL (TGRL) language. We evaluate the approach and tool using a GRL model describing the objectives of a Virtual Interior Designer application along with a set of 42 requirements addressing both security and non-security aspects. The approach and tool yielded positive results, demonstrating a precision of 100%, a recall of 78.5%, and an F1-score of 87.9%.