LLM Application Security Controls

This unit covers operational security controls for Microsoft Azure OpenAI and LLM-based applications, emphasizing LLM-specific risks (OWASP Top 10) and practical controls such as access control, audit logging, network isolation, encryption at rest, and content filtering. It builds on prior foundational concepts and prepares students to design, implement, and validate secure runtime environments and compliance controls in later units.

Learning Objectives

• Analyze the OWASP Top 10 threats for LLM applications and prioritize mitigations relevant to a Microsoft Azure OpenAI deployment
• Apply role-based access control, managed identities, and token/secret management techniques to enforce least-privilege access for Azure OpenAI resources
• Demonstrate the configuration of audit logging and monitoring (e.g., Azure Monitor/Azure Sentinel) and show how logs capture LLM requests, system events, and policy violations for forensic and compliance purposes
• Evaluate and design network isolation, encryption-at-rest strategies, and content-filtering controls (including policy, model, and pipeline-level filters) to protect sensitive data and reduce emergent risk in production LLM services