Masterarbeit (M.Sc.) im Studiengang IT-Sicherheit von Kasimir Romer: Building a Data Processing Pipeline for Submarine Cable News – Leveraging Large Language Models for Advanced Analysis and Visualization

Abstract: Submarine Data Cables (SDCs) are critical infrastructures that handle the majority of global network traffic. Despite their importance, SDCs receive limited media coverage and public attention, which creates significant gaps in understanding their vulnerabilities and significance. This thesis explores the potential of Large Language Models (LLMs) in extracting information about Submarine Data Cables from unstructured text, such as news articles, and investigates methods to improve this extraction process. An LLM-based information extraction module is developed and integrated into the Submarine Cable Dashboard to enhance its ability to process and analyze relevant news data. The performance of this LLM-based module is then compared to the existing rule-based approach. Additionally, the thesis examines and compares different LLMs with respect to accuracy, hallucination rate, and processing speed to determine the most effective model for this specific task. Results show that the LLM-based approach outperforms the rule-based method in most cases, particularly due to the strengths of LLMs in Natural Language Processing tasks.

 Betreuer: Jonas Franken

Prüfer: Prof. Dr. Dr. Christian Reuter