This program addresses the security challenges in AI systems, focusing on adversarial attacks, defense mechanisms, and robust AI design. We explore how to make machine learning models resilient against manipulation and exploitation.
Key Topics Covered
- Adversarial Attacks: FGSM, PGD, and physical-world attacks on vision models.
- Defense Mechanisms: Adversarial training, detection, and robust optimization.
- AI Privacy: Federated learning, differential privacy, and secure multi-party computation.
- Explainable AI (XAI): Techniques to interpret and audit AI decisions.
Learning Outcomes
- Identify and mitigate adversarial vulnerabilities in deep learning models.
- Implement secure AI systems for critical applications (e.g., healthcare, finance).
- Design privacy-preserving machine learning pipelines.
Who Should Join?
- AI security researchers and cybersecurity professionals.
- Engineers working on high-stakes AI applications (autonomous systems, healthcare).
- Policymakers and auditors interested in AI governance and ethics.
Format
- Workshops: Hands-on adversarial attack and defense labs.
- Case Studies: Analyze real-world AI security breaches and mitigation strategies.
- Projects: Build an adversarial detection system or secure federated learning framework.