Abstract
dc:description.abstractMany organizations maintain internal domain names and URLs that are private and not publicly accessible. This thesis explores the challenge of predicting such unknown or private URLs using modern Large Language Models and Deep Learning techniques. The task is particularly difficult due to the lack of visibility into the types and structures of URLs stored in local DNS servers. While Large Language Models (LLMs) have shown impressive capabilities across various domains, they are prone to hallucination and may struggle with generating realistic, previously unseen URLs. In this work, we investigate the use of LLMs—either through training from scratch or fine-tuning pre-trained models—to generate and predict valid domain names. Our approach leverages a large corpus of existing global and organizational DNS data to guide the model toward plausible and contextually appropriate URL predictions.
Degree
thesis:*- Name thesis:degree_name
- Master of Science
- Level thesis:degree_level
- masters
- Discipline thesis:degree_discipline
- Computer Science
- Grantor
- Boston University
- Year dc:date.issued
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Sharma, Ayush
- Advisor dc:contributor.advisor
-
- Crovella, Mark
Subjects
dc:subject × 1Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/2144/53423