New PassGPT+ model enhances password security analysis with linguistic priors

Researchers have developed PassGPT+, a new model that leverages linguistic priors from GPT-2 to improve password modeling. This approach adapts the language model's understanding of language structure to the specific patterns found in human-generated passwords. The study also introduced PassDiffusion, a discrete diffusion model for password generation, which was found to be significantly less effective than autoregressive models for this task. PassGPT+ demonstrated a 16% relative improvement over its predecessor, PassGPT, on the RockYou benchmark, and showed strong generalization to a more recent dataset. AI

IMPACT Improves security analysis by better modeling human password generation patterns.

RANK_REASON Academic paper detailing a new model and its performance. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• GPT-2
• PassDiffusion
• PassGPT+
• RockYou

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →