Securing AI Systems & Data
Overview
AI systems create new targets for attackers and new ways for data to leak. This course explains the security threats specific to AI, from manipulated prompts and poisoned data to model theft and unsafe integrations, and gives participants the controls and practices needed to deploy AI securely.
Who should attend
Information security managers, IT and risk professionals, data protection officers, and architects of AI solutions.
Training methodology
Interactive presentations, live demonstrations of AI tools, hands-on exercises using participants' own work examples, case studies, group discussion, and a personal action plan.
Objectives
By the end of this course, participants will be able to:
- Identify security threats specific to AI systems
- Protect sensitive data used with AI tools
- Assess the security of AI vendors and integrations
- Apply secure design and testing practices
- Respond to AI-related security incidents
Course outline
Day 1The AI Threat Landscape
- How AI changes the attack surface
- Prompt injection and manipulation
- Data poisoning and model theft
- Real-world incidents
- Deepfakes and social engineering
- Insider risks
Day 2Protecting Data
- Data classification for AI use
- Leakage through prompts and outputs
- Access control and encryption
- Policy exercise
- Approved tools and usage rules
- Monitoring data flows
Day 3Vendors and Integrations
- Security due diligence
- Plugins, APIs and agents
- Contractual safeguards
- Vendor assessment
- Supply chain risk in AI models
- Cloud security basics
Day 4Secure Design and Testing
- Security by design for AI
- Red-teaming and adversarial testing
- Logging and monitoring
- Testing workshop
- Security frameworks for AI
- Penetration testing scope
Day 5Incident Response
- AI incident scenarios
- Containment and recovery
- Lessons learned
- Security action plan
- Communication during incidents
- Tabletop exercise
