Adversarial AI & ML Security
Overview
Adversarial AI & ML Security is an intensive, hands-on course that teaches you how AI systems fail—and how to make them resilient. You’ll explore attacks across the lifecycle (data collection, training, fine-tuning, retrieval, inference, and agent/tool use), including data poisoning/backdoors, adversarial examples, model and data confidentiality attacks, and application-layer threats like prompt injection and tool abuse in RAG and multi-agent systems. You’ll implement layered defenses: input validation and content filters, retrieval sanitization, capability and tool-use controls, sandboxing, adversarial training, privacy protections, and operational guardrails with monitoring and incident playbooks. By the end, you’ll be able to evaluate and understand how to harden LLM and classical ML solutions with security-by-design principles.
Duration
2 Days
Who Should Take This Course
Audience
Security Engineers/Architects, ML/LLM Engineers, Developers, SRE/Platform, Product & Risk/Compliance.
Prerequisites
Participants should have basic Python skills and familiarity with AI/ML concepts. Prior exposure to application security or cloud/security tooling is helpful but not required.
Why You Should Take This Course
Upon completing Adversarial AI & ML Security, participants will be able to:
Upon completing Adversarial AI & ML Security, attendees will be able to:
- Identify the AI/ML threat landscape and map risks to systems, data, and users
- Explain and demonstrate common attacks (poisoning, evasion, model extraction, membership inference, prompt injection/jailbreaks)
- Threat-model LLM/RAG/agentic applications and data pipelines end-to-end
- Implement practical mitigations (guardrails, policy enforcement, sandboxing, rate-limiting, data hygiene, adversarial training, privacy techniques)
- Design secure architectures for model training, serving, and tool use (RAG/agents)
- Establish red-teaming, evaluation, and monitoring practices for ongoing resilience
- Align controls with organizational governance and AI risk frameworks
Course Outline
Adversarial AI & ML Security
- AI/ML Threat Landscape & Taxonomy
- Data Supply Chain & Poisoning Attacks
- Evasion & Adversarial Examples
- Model & Data Confidentiality Attacks
- Securing LLM/RAG/Agent Applications
- Defensive Techniques & Design Patterns
- Evaluation, Red Teaming & Monitoring
- Governance & Secure Architecture for Deployment