Luisa Crawford
Sep 19, 2024 10:04
NVIDIA highlights AI security advancements at Black Hat USA and DEF CON 32, emphasizing adversarial machine learning and LLM security.
NVIDIA recently demonstrated its AI security expertise at two of the most prestigious cybersecurity conferences, Black Hat USA and DEF CON 32, according to the NVIDIA Technical Blog. The events provided a platform for NVIDIA to showcase its latest advancements in AI security and share insights with the broader cybersecurity community.
NVIDIA at Black Hat USA 2024
The Black Hat USA conference is a globally recognized event that features cutting-edge security research. This year, discussions highlighted the applications of generative AI tools in security and the security of AI deployments. Bartley Richardson, NVIDIA’s Director of Cybersecurity AI, delivered a keynote alongside WWT CEO Jim Kavanaugh, focusing on how AI and automation are transforming cybersecurity strategies.
Other sessions featured experts from NVIDIA and its partners discussing the revolutionary impact of AI on security postures and techniques for securing AI systems. A panel on AI Safety included Nikki Pope, NVIDIA’s Senior Director of AI and Legal Ethics, who discussed the complexities of AI safety with practitioners from Microsoft and Google.
Daniel Rohrer, NVIDIA’s VP of Software Product Security, addressed the unique challenges of securing AI data centers in a session hosted by Trend Micro. The consensus at Black Hat was clear: deploying AI tools necessitates a robust approach to security, emphasizing trust boundaries and access controls.
NVIDIA at DEF CON 32
DEF CON, the world’s largest hacker conference, featured numerous villages where attendees engaged in real-time hacking challenges. NVIDIA researchers supported the AI Village, hosting popular live red-teaming events focused on large language models (LLMs). This year’s events included a Generative Red Team challenge, which led to real-time improvements in model safety guardrails.
Nikki Pope delivered a keynote on algorithmic fairness and safety in AI systems. The AI Cyber Challenge (AIxCC), hosted by DARPA, saw red and blue teams building autonomous agents to identify and exploit code vulnerabilities. This initiative underscored the potential of AI-powered tools to accelerate security research.
Adversarial Machine Learning Training
At Black Hat, NVIDIA and Dreadnode conducted a two-day training on machine learning (ML), covering techniques to assess security risks against ML models and implement specific attacks. Topics included evasion, extraction, assessments, inversion, poisoning, and attacks on LLMs. Participants practiced executing these attacks in self-paced labs, gaining hands-on experience critical for shaping effective defensive strategies.
Focus on LLM Security
NVIDIA Principal Security Architect Rich Harang presented on LLM security at Black Hat, emphasizing the importance of grounding LLM security in a familiar application security framework. The talk focused on the security issues associated with retrieval-augmented generation (RAG) LLM architectures, which significantly expand the attack surface of AI models.
Attendees were advised to identify and analyze trust and security boundaries, trace data flows, and apply the principles of least privilege and output minimization to ensure robust security.
Democratizing LLM Security Assessments
At DEF CON, NVIDIA AI Security Researchers Leon Derczynski and Erick Galinkin introduced garak, an open-source tool for LLM security probing. Garak allows practitioners to test potential LLM exploits quickly, automating a portion of LLM red-teaming. The tool supports nearly 120 unique attack probes, including XSS attacks, prompt injection, and safety jailbreaks.
Garak’s presentation and demo lab were well-attended, marking a significant step forward in standardizing security definitions for LLMs. The tool is available on GitHub, enabling researchers and developers to quantify and compare model security against various attacks.
Summary
NVIDIA’s participation in Black Hat USA and DEF CON 32 highlighted its commitment to advancing AI security. The company’s contributions provided the security community with valuable knowledge for deploying AI systems with a security mindset. For those interested in adversarial machine learning, NVIDIA offers a self-paced online course through its Deep Learning Institute.
For more insights into NVIDIA’s ongoing work in AI and cybersecurity, visit the NVIDIA Technical Blog.
Image source: Shutterstock
