Ημερομηνία: Τετάρτη 18/9 στις 13:00
Φοιτήτρια: Ραφαήλα Γαλανοπούλου
Τίτλος: Exploring the Power of Data: A Deep Dive into Large Language Models for Vulnerability Detection
Σύνδεσμος: https://meet.google.com/wfm-jrtd-maq
Περίληψη:
The increasing prevalence of software vulnerabilities and the advancements in Natural Language Processing (NLP) have paved the way for exploring the potential of Large Language Models (LLMs) in vulnerability detection. This thesis investigates the impact of datasets on the performance of LLMs in identifying vulnerabilities within source code.
We delve into the correlation between model efficacy and code complexity metrics, dataset structure, and the specific types of vulnerabilities.
Our findings highlight the significant influence of dataset characteristics on the performance of LLMs, emphasizing the need for tailored training data and fine-tuning strategies. We also explore the potential of incorporating a data evaluation step during preprocessing to measure factors like code similarity, which could further enhance the model’s effectiveness. This research contributes to the ongoing efforts in leveraging LLMs for improving software security and provides insights for future research directions in this domain.
Η ομιλία αυτή αποτελεί τη δημόσια παρουσίαση της διπλωματικής εργασίας της ομιλήτριας.