Home » Courses » Certification Courses » AI Certificates » Certified Retrieval-Augmented Generation Security Fundamentals (CRAGSF)

Certified Retrieval-Augmented Generation Security Fundamentals (CRAGSF)

Length: 2 Days
Print Friendly, PDF & Email

Certified Retrieval-Augmented Generation Security Fundamentals (CRAGSF) Certification Program by Tonex

Certified Retrieval-Augmented Generation Security Fundamentals (CRAGSF)

Duration: 2 Days | Format: Virtual / In-Person / Hybrid
Credential: Certification + Exam + Digital Badge
Level: Foundation – Technical Awareness

The CRAGSF certification equips AI engineers, cybersecurity professionals, architects, and compliance teams with foundational knowledge of security risks, vulnerabilities, and mitigation strategies for Retrieval-Augmented Generation (RAG) systems. RAG blends LLMs with external knowledge bases or vector databases, introducing novel attack surfaces and integrity risks. This course focuses on real-world RAG implementations, attack modeling, and risk management.

Learning Objectives

By completing the CRAGSF program, participants will be able to:

  • Describe the architecture and components of RAG-based systems.
  • Identify major security risks in RAG pipelines, including prompt injection, data poisoning, and vector store exploits.
  • Analyze security implications of document retrieval, embedding strategies, and indexing.
  • Map known threats using OWASP LLM Top 10, MITRE ATLAS, and AI-specific threat modeling tools.
  • Apply secure RAG design principles to prevent data leakage, hallucination amplification, and unauthorized access.
  • Develop a basic RAG-specific risk register and mitigation plan for LLM-integrated systems.
  • Support secure MLOps pipelines and compliance readiness for RAG-based deployments.

Target Audience:

  • AI/ML Engineers & LLM Developers
  • Security Architects and Red Teamers
  • DevSecOps & MLOps Professionals
  • Application and Data Security Engineers
  • Compliance & Risk Managers
  • Product Managers working with GenAI or LLM applications
  • Technical Team Leads and CISOs adopting LLMs

Program Modules:

Day 1 – RAG Fundamentals and Threat Landscape

Module 1: Introduction to RAG Architecture

  • Components: Retriever, Ranker, Embedder, Vector Store, Generator
  • Contrast with vanilla LLMs and fine-tuned models
  • Use cases: knowledge agents, semantic search bots, copilots

Module 2: Threat Landscape for RAG Systems

  • OWASP Top 10 for LLMs (2024)
  • MITRE ATLAS threat actor behaviors
  • Prompt injection (direct, indirect, cross-layer)
  • Embedding layer manipulation and injection

Module 3: Secure Retrieval Layer Engineering

  • Data sanitization and content-based access control
  • Vector store attacks: collision, poisoning, inference
  • Secure chunking, indexing, and metadata control
  • Case: Document retrieval manipulation via pre-crafted text blocks

Module 4: Input/Output Risk and Jailbreak Defense

  • Generator-level risks (e.g., hallucination escalation, sensitive echo)
  • RAG-specific prompt hardening
  • Context window exploitation and contextual injections
  • Output validation and restricted formatting

Day 2 – Defensive Strategies and Secure RAG Implementation

Module 5: Security Controls and Best Practices

  • RAG pipeline authentication and encryption
  • Separation of concerns: isolate retriever and generator
  • Content moderation of retrieved chunks
  • RAG + RBAC (Role-based chunk access)

Module 6: RAG Risk Management and Governance

  • RAG attack tree design (with DFDs and STRIDE extensions)
  • Building a RAG-specific risk register
  • Integration with AI RMF, NIST AI 100-1, ISO 42001
  • Logging, audit trails, and monitoring for RAG interactions

Module 7: Secure RAG Development Lifecycle

  • Secure MLOps for RAG systems
  • Model card extensions for RAG
  • CI/CD for vector stores and retrieval functions
  • Pen testing & red teaming RAG pipelines

Module 8: Final Capstone Simulation

  • Use case: Secure RAG assistant for enterprise documentation
  • Teams define threat model, secure architecture, and test plan
  • Walkthrough: data leak prevention, zero trust retrieval, and prompt control

Course Materials:

  • CRAGSF slide deck ( PDF)
  • Participant workbook with diagrams, templates, and quizzes
  • RAG security checklist (architecture + runtime)
  • Threat modeling templates (DFD, STRIDE, OWASP, MITRE ATT@CK and ATLAS)
  • Secure embedding pipeline design cheat sheet
  • Prompt injection test cases & benchmarking scripts (optional)

Certification Exam:

  • Format: 50 questions (MCQ, scenario-based)
  • Passing Score: 70%
  • Duration: 90 minutes
  • Credential: Certified RAG Security Fundamentals (CRAGSF)
  • Badge Validity: 1 year (renewable via CEUs or re-exam)
  • Badge Platform: Badge.ink / Open Badges compatible

 CRAGSF Certification Exam Domains

Domain Weight Description
1. RAG Architecture & Components 15% Understand the structure and workflow of RAG systems including the retriever, embedder, vector store, and generator. Covers how data is ingested, stored, and retrieved for generation.
2. Threat Landscape for RAG Systems 20% Covers known and emerging security threats such as prompt injection (direct/indirect), data poisoning, embedding attacks, and vector database exploits. Maps OWASP Top 10 for LLMs and MITRE ATLAS.
3. Retrieval Layer and Vector Store Security 15% Focuses on securing document ingestion, chunking, indexing, and access control. Includes protection against inference attacks, embedding manipulation, and access-based filtering.
4. Prompt and Output Hardening Techniques 15% Techniques to mitigate prompt leakage, jailbreaking, and output manipulation. Includes input validation, output filtering, and role-constrained generation.
5. Secure Development and MLOps Practices for RAG 15% Integration of RAG systems into secure MLOps pipelines, CI/CD for retrievers and generators, secure embedding models, and model lifecycle management.
6. Risk Management, Governance & Compliance 10% Includes risk register development, threat modeling, compliance alignment (NIST AI RMF, ISO 42001), audit trails, and governance policies for RAG pipelines.
7. Secure Deployment and Monitoring of RAG Systems 10% Covers logging, access control, anomaly detection, runtime validation, and use of secure APIs in deployed RAG environments.

Passing Criteria

  • Exam Format: 50–60 multiple choice and scenario-based questions
  • Duration: 90 minutes
  • Passing Score: 70%
  • Delivery: Online or proctored exam platform
  • Credential Awarded: Certified RAG Security Fundamentals (CRAGSF)

 

Request More Information