Open-weight LLMs evaluated for autonomous vehicle threat intelligence generation

Researchers have developed a new dataset, CAV-STIXGen, to evaluate open-weight Large Language Models (LLMs) in generating structured threat information for autonomous vehicle vulnerabilities. The study assessed 11 LLMs, with single-model configurations achieving high F1 scores for structured data objects (SDO) and Common Weakness Enumeration (CWE) mapping, though MITRE ATT&CK mapping proved more difficult. A multi-agent setup using Gemma-4-31B and Codestral-22B also showed promising results for SDO and SRO generation, indicating AI's potential to automate threat intelligence in transportation security. AI

IMPACT Automates threat intelligence generation for autonomous vehicle security, potentially improving defense prioritization.

RANK_REASON The cluster contains an academic paper detailing research on LLM capabilities for a specific security domain. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• CAV-STIXGen
• Codestral-22B
• Common Weakness Enumeration
• Gemma-4-31B
• Large Language Models
• MITRE ATT&CK
• Structured Threat Information Expression

• paper
• product

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