Security-Focused Chaos Engineering - The Lasso for AI Security Threats

Explore security-focused chaos engineering as a proactive approach to addressing AI security threats in this 29-minute conference talk. Learn how to apply chaos testing principles to AI security using the Secops-Chaos open-source framework. Discover techniques for encoding Tactics, Techniques, and Procedures (TTPs) as security-focused chaos experiments, with practical demonstrations mapping MITRE ATLAS TTPs to AI applications running in Kubernetes environments. Gain insights into evolving threat vectors in AI applications, understand the limitations of existing frameworks like OWASP LLM risks and MITRE ATLAS, and learn how to customize these baseline frameworks for your organization's specific needs. Delve into the importance of continuous verification for AI applications, particularly when dealing with indeterministic behavior and third-party models. Equip yourself with knowledge to actively test and improve the security of organizational data, intellectual property, and internal APIs connected through AI applications.