Course Objective & Scope
RAG systems connected to enterprise knowledge bases and AI agents capable of invoking external tools introduce risks beyond traditional AppSec: indirect prompt injection, unauthorised tool execution, cross-tenant vector leakage and unbounded autonomous actions. Aligned with the OWASP Top 10 for LLM and Agentic Applications, this course covers securing LLM, RAG and multi-agent architectures.
In a controlled lab, participants red-team a sample LLM/agent application and harden it using input/output guardrails, document-level ACLs in retrieval, MCP/tool authorization boundaries and sandboxed execution.
Who Should Attend?
- AI and software engineers building LLM, RAG and AI agent applications
- Application security (AppSec), AI Red Team and cloud security specialists
- Enterprise AI platform architects and technical leads
- Prerequisite: experience with LLM APIs, prompt/RAG architectures and web/API development
Key Takeaways
- Assess LLM and agent architectures against the OWASP Top 10 for LLM Applications
- Build defense-in-depth against direct and indirect prompt injection and jailbreak attempts
- Enforce document/chunk-level ACLs and PII redaction across RAG pipelines
- Implement least privilege, schema validation, sandboxing and human approval for tool calling and MCP servers
- Embed automated AI red teaming and security eval checklists into CI/CD workflows
Curriculum
LLM and Agent Attack Surface (OWASP Top 10 for LLM)
- Prompt injection, sensitive information disclosure, insecure output handling and excessive agency
- Model supply chain: open-weight file formats (safetensors vs pickle) and adapter security
Direct / Indirect Prompt Injection and Guardrail Architecture
- Indirect prompt injection delivered through documents, web pages and emails
- Instruction/data separation, input/output classifiers, PII redaction and safe output parsing
RAG and Vector Database Security
- Permission-aware retrieval (document-level ACLs) and multi-tenant isolation
- Mitigating knowledge base poisoning and embedding leakage
Agent, Tool Calling and MCP Security
- On-behalf-of identity propagation and scope restriction in tool calling and Model Context Protocol (MCP)
- Code-execution sandboxes, loop/budget caps and human-in-the-loop approval for sensitive actions
AI Red Teaming and Continuous Security Validation
- Automated adversarial testing with tools such as Promptfoo, Garak and PyRIT
- Pre-production AI & Agent security checklist and runtime observability/tracing
Other Courses in This Category
-
ISO/IEC 27001:2022 clauses, asset and risk assessment, Annex A controls, Statement of Applicability (SoA) and internal audit readiness for teams establishing and operating an ISMS.
- Live Course
- 5 modules
-
The Linux network stack, nftables/iptables, SSH and privilege hardening, SELinux/AppArmor, auditd and packet analysis for systems and security engineers securing production Linux servers.
- Live Course
- 5 modules
-
Static and dynamic malware analysis in isolated labs, PE/ELF inspection, behavioural monitoring, unpacking, code analysis with Ghidra and YARA rule authoring for security analysts.
- Live Course
- 5 modules
-
All 8 domains of the (ISC)² CISSP Common Body of Knowledge (CBK), risk and governance mindset, security architecture and scenario-based question analysis for experienced security professionals.
- Live Course
- 5 modules
-
The CIA triad, modern threat vectors, identity and access management, network security and baseline hardening practices for IT and software teams who need to identify and mitigate enterprise security risks.
- Live Course
- 5 modules
-
Identity, device posture, network micro-segmentation and continuous verification using NIST SP 800-207 and the CISA Zero Trust Maturity Model for infrastructure and security architects planning enterprise adoption.
- Live Course
- 5 modules
Get in touch
