Adversarial AI and Threat Modeling

This unit introduces common attack surfaces, threat modeling techniques, and data-security practices specific to AI and machine learning systems. Students explore adversarial examples, privacy risks, and baseline hardening strategies to prepare for hands-on defenses in subsequent units.

Learning Objectives

• Analyze AI/ML system architectures to identify attack surfaces, threat actors, and likely adversarial goals using a structured threat-modeling approach
• Apply secure data-handling and privacy-preserving techniques (e.g., data minimization, anonymization, access controls) to reduce risk in model training and deployment pipelines
• Demonstrate generation and detection of simple adversarial examples in a controlled lab setting and document their impact on model performance
• Evaluate model-hardening and monitoring strategies (e.g., input validation, ensemble methods, logging/alerting) and justify trade-offs between security, accuracy, and performance