Sécuriser les systèmes d'IA à toutes les étapes du cycle de vie

Aperçu

As artificial intelligence powers our world, it creates a new frontier for complex threats that standard cybersecurity practices can't handle. This course equips you with the specialized, in-demand skills to defend these critical systems from end to end.

You will learn to think like an attacker, identifying unique threats like data poisoning, adversarial evasion, and model inference attacks. We'll journey through the entire MLOps lifecycle, pinpointing vulnerabilities from the moment data is collected to the second a model is deployed.

But this isn't just theory—you will immediately apply your knowledge in a series of hands-on labs. Using the industry-standard MITRE ATLAS framework, you'll perform a full threat model analysis on a sample AI application.

You will then implement practical, code-based mitigation strategies to build more resilient systems, culminating your learning in a final project where you conduct a full security audit. This course is ideal for AI engineers, data scientists, cybersecurity professionals, and anyone involved in the design, development, or deployment of AI systems.

It is especially valuable for professionals working in sectors where security is a priority, such as healthcare, finance, and government. Learners should have a foundational understanding of AI, machine learning, and basic cybersecurity concepts.

Familiarity with software development practices and system architecture will be beneficial, but not required. By the end of this course, you will have the confidence and tangible skills to protect the next generation of technology and become an essential asset in the world of AI security.

Programme

• Le paysage des menaces de l'IA
Ce module introduit les apprenants au paysage de la sécurité de l'IA. Il décompose les principales catégories d'attaques visant les systèmes d'IA et introduit des cadres fondamentaux pour comprendre et classer ces menaces émergentes.
• Construction et suivi de modèles avec MLflow
Ce module se concentre sur l'utilisation de MLflow pour le suivi des expériences et la gestion des modèles, un élément crucial de MLOps sur Databricks.
• Déploiement et gestion des modèles
Ce module conclut le cycle de vie ML en couvrant le déploiement et la gestion des modèles en utilisant le MLflow Model Registry.

---

Enseigné par

Ashish Mohan and Starweaver

---

Matières

Computer Science