AI models generate PowerShell malware with high similarity to real-world samples

Researchers have developed an experimental framework to assess the capabilities of large language models (LLMs) in generating PowerShell malware. This framework includes a novel sandbox approach for dynamic analysis and a curated dataset of real-world PowerShell malware. The study found that permissive, open-weight LLMs can generate malware highly similar to human-written samples, with a median Jaccard index of 84.5% and nearly half of generated instances showing complete overlap with real malware. AI

IMPACT Highlights the growing threat of AI-generated malware, necessitating advanced detection and analysis techniques for cybersecurity professionals.

RANK_REASON The cluster is based on a research paper published on arXiv detailing an experimental framework and dataset for analyzing AI-generated malware. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• AI Code Generators for Security: Friend or Foe?
• arXiv
• dataset
• Experimental framework and evaluation of the 5G-Crosshaul control infrastructure
• generative artificial intelligence
• Jaccard index
• Open-weight LLMs
• OS malicious events
• PowerShell
• sandbox approach

• paper
• safety

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