Course Outline
Overview of Artificial Intelligence Threat Modeling
- Factors contributing to vulnerabilities in artificial intelligence systems
- Comparison of attack surfaces between AI applications and traditional information technology systems
- Primary attack vectors encompassing data, model parameters, output mechanisms, and interface points
Adversarial Attacks on Artificial Intelligence Models
- Analysis of adversarial examples and perturbation methodologies
- Distinctions between white-box and black-box attack frameworks
- Examination of Fast Gradient Sign Method (FGSM), Projected Gradient Descent (PGD), and DeepFool techniques
- Methods for visualizing and generating adversarial samples
Model Inversion and Privacy Leakage Risks
- Techniques for infencing training data from model outputs
- Membership inference attacks and their implications
- Privacy vulnerabilities associated with classification and generative models
Data Poisoning and Backdoor Injection Mechanisms
- Impact of corrupted data on model performance and behavior
- Trigger-based backdoors and Trojan attack implementations
- Strategies for detection and data sanitization
Robustness and Defense Techniques
- Adversarial training approaches and data augmentation strategies
- Gradient masking and input preprocessing methods
- Model smoothing and regularization techniques
Privacy-Preserving Artificial Intelligence Defenses
- Foundations of differential privacy frameworks
- Noise injection mechanisms and management of privacy budgets
- Federated learning protocols and secure aggregation processes
Implementation of Artificial Intelligence Security in Practice
- Threat-aware evaluation and deployment procedures for government systems
- Application of the Adversarial Robustness Toolbox (ART) in operational environments
- Industry case studies detailing real-world security incidents and mitigation strategies
Executive Summary and Next Steps
Requirements
- Proficiency in machine learning operational processes and algorithm development cycles
- Demonstrated expertise in Python programming and standard machine learning libraries, including PyTorch or TensorFlow
- Knowledge of foundational security principles and threat assessment methodologies is advantageous
Target Audience
- Machine learning engineers
- Cybersecurity analysts
- AI researchers and model validation teams engaged in governance and compliance for government initiatives
Testimonials (2)
I really enjoyed learning about AI attacks and the tools out there to begin practicing and actively using for security testing. I took a lot of knowledge away which I didn't have at the beginning and the course met what I hoped it would be. My favorite part shown from the training was Comet Browser and was amazed at what it could do. Definitely something will be looking into more. Overall it was a great course and enjoyed learning all OWASP GenAI Top 10.
Patrick Collins - Optum
Course - OWASP GenAI Security
The profesional knolage and the way how he presented it before us
