Get in Touch

Course Outline

Introduction to AI Red Teaming

  • Navigating the evolving AI threat landscape
  • The critical role of red teams in AI security
  • Key ethical and legal implications

Adversarial Machine Learning

  • Attack vectors: evasion, poisoning, model extraction, and inference attacks
  • Creation of adversarial examples using techniques like FGSM and PGD
  • Differentiating targeted vs. untargeted attacks and defining success metrics

Evaluating Model Robustness

  • Measuring robustness against various perturbations
  • Identifying model blind spots and potential failure modes
  • Stress testing classification, vision, and NLP models

Red Teaming AI Pipelines

  • Mapping the attack surface across data, model, and deployment layers
  • Exploiting insecure model APIs and endpoints
  • Reverse engineering to understand model behavior and outputs

Simulation and Tooling

  • Utilizing the Adversarial Robustness Toolbox (ART)
  • Employing tools such as TextAttack and IBM ART for red teaming
  • Leveraging sandboxing, monitoring, and observability tools

AI Red Team Strategy and Defense Collaboration

  • Structuring red team exercises and defining clear objectives
  • Effectively communicating findings to blue teams
  • Embedding red teaming practices into broader AI risk management

Summary and Recommended Next Steps

Requirements

  • A solid grasp of machine learning and deep learning architectures
  • Proficiency in Python and major ML frameworks (e.g., TensorFlow, PyTorch)
  • Knowledge of cybersecurity principles or offensive security methodologies

Target Audience

  • Security researchers
  • Offensive security teams
  • AI assurance and red team specialists
 14 Hours

Number of participants


Price per participant

Testimonials (3)

Upcoming Courses

Related Categories