New framework helps investigators bypass dark web CAPTCHAs

Researchers have developed DarKnight, an agentic framework designed to help cyber threat intelligence investigators bypass CAPTCHAs on dark web marketplaces. This framework utilizes locally hosted, open-weight vision-language models (VLMs) and a retrieval-augmented generation (RAG) approach to improve CAPTCHA-solving capabilities. In a case study, DarKnight achieved an 83.2% success rate on a heavily defended market, significantly outperforming its base model and even the advanced Claude Opus model. AI

IMPACT This framework could enhance the efficiency and autonomy of cybersecurity investigations on the dark web.

RANK_REASON The cluster describes a new research framework and its performance on a specific task, published on arXiv. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• arXiv
• Claude Opus
• cyber threat intelligence
• Hugging Face
• Levente Csikor PhD
• Tor
• vision-language model

• other

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