LLM Security Controls and Monitoring

This unit examines threats specific to large language model applications and teaches technical controls to harden Azure OpenAI deployments. Students will learn OWASP LLM risks and implement access control, audit logging, network isolation, encryption-at-rest, and content filtering to prepare for secure deployment and incident response work in subsequent units.

Learning Objectives

• Analyze the OWASP Top 10 for LLM applications and map at least five specific vulnerabilities to plausible attack scenarios in an Azure OpenAI environment.
• Apply role-based and least-privilege access control principles to design and implement access policies and key scoping for Azure OpenAI resources in a lab configuration.
• Demonstrate configuration of audit logging, network isolation, and encryption-at-rest controls by generating, collecting, and interpreting log events and showing encrypted storage for model data in a simulated incident exercise.
• Evaluate content-filtering and mitigation strategies (e.g., prompt sanitization, input/output filtering, rate limiting) and produce a prioritized mitigation plan addressing privacy, data exfiltration, and misuse risks.