Securing AI Models: Threats, Attacks, and Defenses Training Course

Introduction to AI Threat Modeling for Government

• Factors contributing to the vulnerability of AI systems in government operations
• Comparison of AI attack surfaces with traditional IT systems
• Key attack vectors: data, model, output, and interface layers for government applications

Adversarial Attacks on AI Models for Government

• Understanding adversarial examples and perturbation techniques in the context of government systems
• Differentiating between white-box and black-box attacks in government AI models
• Overview of FGSM, PGD, and DeepFool methods for government applications
• Techniques for visualizing and crafting adversarial samples for government use cases

Model Inversion and Privacy Leakage for Government

• Methods for inferring training data from model outputs in government systems
• Analysis of membership inference attacks on government AI models
• Evaluation of privacy risks in classification and generative models used by government agencies

Data Poisoning and Backdoor Injections for Government

• Impact of poisoned data on model behavior in government applications
• Strategies for trigger-based backdoors and Trojan attacks in government AI systems
• Detection and sanitization strategies to protect government AI models from data poisoning

Robustness and Defense Techniques for Government

• Implementation of adversarial training and data augmentation in government AI systems
• Use of gradient masking and input preprocessing techniques to enhance security in government models
• Application of model smoothing and regularization techniques for robustness in government AI

Privacy-Preserving AI Defenses for Government

• Introduction to differential privacy in government data management
• Techniques for noise injection and managing privacy budgets in government systems
• Utilization of federated learning and secure aggregation methods for government AI security

AI Security in Practice for Government

• Threat-aware model evaluation and deployment strategies for government agencies
• Application of the Adversarial Robustness Toolbox (ART) in government settings
• Industry case studies: real-world breaches and mitigations relevant to government operations

Summary and Next Steps for Government