Unforgettable Password Generation Using LoRA Fine-Tuned Large Language Model

In response to the challenges posed by complex and hard-to-remember passwords in traditional authentication systems, this research introduces an innovative approach leveraging Low Rank Adaption (LoRA) fine-tuned Large Language Models (LLMs) to generate natural language-like strong passwords. Focused on striking a balance between usability and security, proposed system methodology employs smaller-sized pre-trained LLMs, adapted using LoRA fine-tuning, to produce passwords that are both strong and memorable. By addressing the limitations of current password generation practices, proposed system framework aims to encourage user adoption of secure practices, contributing to a positive societal impact through improved cybersecurity hygiene. The proposed solution involves deploying a fine-tuned LLM and serving them through an frontend web UI which emphasizes resource efficiency scalability of deployment of such model, and ease of using the generated passwords in real life authentication purposes fostering good cybersecurity practices among non-technical users.