Model Red-Teaming: Dynamic Security Analysis for LLMs
Date: Thursday, October 23rd at 8am PT / 11am ET
The rise of Large Language Models has many organizations rushing to integrate AI-powered tools into existing products, but they introduce significant new risk. OWASP has recently introduced the LLM Top 10 to highlight these novel threat vectors, including prompt injection and data exfiltration. However, existing AppSec tools are not designed to detect and remediate these vulnerabilities. In particular, static analysis (SAST), one of the most common tools, cannot be used since there is no code: machine-learning models are effectively “black boxes."
LLM red-teaming is emerging as a technique to minimize the vulnerabilities associated with LLM adoption, ensure data confidentiality, and verify that safety and ethical guardrails are being applied. It applies tactics associated with penetration testing and dynamic analysis (DAST) of traditional software to the new world of machine-learning models.
Join Snyk, the leader in AI Trust, for an overview of LLM red-teaming principles, including:
- What are some of the novel threat vectors associated with large language models, and how are these attacks carried out?
- Why are traditional vulnerability-detection tools (such as SAST and SCA) incapable of detecting the most serious risks in LLMs?
- How can the principles of traditional dynamic analysis be applied to machine learning models, and what types of new tools are needed?
- How should organizations begin to approach building an effective program for LLM red-teaming?
Register now!
About the presenter:
Clinton Herget, Field CTO, Snyk
Clinton Herget is Field CTO at Snyk, the leader in Developer Security, where he focuses on crafting and evangelizing our strategic vision for the evolution of DevSecOps. A seasoned technologist, Clinton spent his 20-year career prior to Snyk as a web software developer, DevOps consultant, cloud solutions architect, and engineering director. Clinton is passionate about empowering software engineers to do their best work in the chaotic cloud-native world, and is a frequent conference speaker, developer advocate, and technical thought leader.