Threat-Informed AI: Mapping Cloud-Native ML Pipelines to the MITRE ATLAS Framework

🤖 Threat-Informed AI: Mapping Cloud-Native ML Pipelines to the MITRE ATLAS Framework
📅 Tuesday, October 27th, 2026
🕖 07:00 CEST / 11:30 IST / 14:00 SGT / 17:00 AEDT
Presenter: Tim Wadhwa-Brown

As organizations rapidly deploy Large Language Models (LLMs) and agentic workflows into production, securing the underlying cloud-native infrastructure becomes a critical priority. Traditional security boundaries are no longer sufficient when vulnerabilities span from Kubernetes orchestrators down to model weights, training data pipelines, and Retrieval-Augmented Generation (RAG) vector databases.

In this session, Tim Wadhwa-Brown will introduce the MITRE ATLAS (Adversarial Threat Landscape for Artificial-Intelligence Systems) framework, exploring how it bridges the gap between traditional threat-informed defense and the unique lifecycle of AI/ML systems. We will deconstruct real-world case studies and map cloud-native AI vulnerabilities—such as prompt injection, data poisoning, and model extraction—directly to ATLAS tactics and techniques. Finally, we will discuss how platform engineers and security teams can leverage tools like the ATLAS Navigator and the Arsenal emulation plugin alongside CNCF-native architectures to build resilient, verifiable, and secure AI platforms.

Key Takeaways for the Audience • Understand the core matrix of MITRE ATLAS and how it expands upon the traditional ATT&CK framework for machine learning. • Identify the unique threat vectors introduced by hosting LLM APIs, vector stores, and training loops on cloud-native infrastructure. • Practical strategies to align MITRE ATLAS with existing cloud-native security controls and governance standards (like the NIST AI RMF).