Thank you for sending your enquiry! One of our team members will contact you shortly.
Thank you for sending your booking! One of our team members will contact you shortly.
Duration 21 hours
Course Outline
Foundations: Threat Models for Agentic AI
- Categorization of agentic threats, including misuse, privilege escalation, data leakage, and supply-chain vulnerabilities.
- Analysis of adversary profiles and attacker capabilities specifically targeting autonomous agents.
- Identification and mapping of key assets, trust boundaries, and critical control points for agent operations.
Governance, Policy, and Risk Management
- Implementation of governance frameworks for agentic systems, defining roles, responsibilities, and approval gates.
- Development of comprehensive policies covering acceptable use, escalation protocols, data handling, and auditability.
- Navigating compliance requirements and strategies for collecting evidence for audits.
Non-Human Identity & Authentication for Agents
- Designing robust identities for agents using service accounts, JWTs, and short-lived credentials.
- Applying least-privilege access patterns and just-in-time credentialing strategies.
- Managing the identity lifecycle, including rotation, delegation, and revocation processes.
Access Controls, Secrets, and Data Protection
- Implementing fine-grained access control models and capability-based patterns for agents.
- Overseeing secrets management, encryption in transit and at rest, and data minimization practices.
- Securing sensitive knowledge sources and PII against unauthorized access by agents.
Observability, Auditing, and Incident Response
- Designing telemetry systems for agent behavior, including intent tracing, command logs, and data provenance.
- Integrating with SIEM solutions, setting alerting thresholds, and ensuring forensic readiness.
- Developing runbooks and playbooks for managing agent-related incidents and containment strategies.
Red-Teaming Agentic Systems
- Planning effective red-team exercises, defining scope, rules of engagement, and safe failover mechanisms.
- Utilizing adversarial techniques such as prompt injection, tool misuse, chain-of-thought manipulation, and API abuse.
- Conducting controlled attacks to measure system exposure and impact.
Hardening and Mitigations
- Deploying engineering controls like response throttles, capability gating, and sandboxing.
- Implementing policy and orchestration controls, including approval flows, human-in-the-loop mechanisms, and governance hooks.
- Applying model and prompt-level defenses through input validation, canonicalization, and output filtering.
Operationalizing Safe Agent Deployments
- Adopting deployment patterns such as staging, canary releases, and progressive rollouts for agents.
- Maintaining rigorous change control, testing pipelines, and pre-deployment safety checks.
- Fostering cross-functional governance through coordinated playbooks involving security, legal, product, and ops teams.
Capstone: Red-Team / Blue-Team Exercise
- Execution of simulated red-team attacks against a sandboxed agent environment.
- Defensive, detection, and remediation activities as the blue team, leveraging controls and telemetry data.
- Presentation of findings, remediation plans, and necessary policy updates.
Summary and Next Steps
Requirements
- A strong foundation in security engineering, system administration, or cloud operations.
- Working knowledge of AI/ML concepts and the behavioral characteristics of large language models (LLMs).
- Practical experience with identity & access management (IAM) and secure system architecture.
Audience
- Security engineers and red-team specialists.
- AI operations and platform engineers.
- Compliance officers and risk management professionals.
- Engineering leads overseeing the deployment of agent-based systems.
Testimonials (1)
inventory and identifying the different risk exposures within AI
