Model Online / On-Site
Duration
Modules 5 modules

Course Objective & Scope

Organisations are rapidly pulling open-source models, datasets and Python libraries into projects, yet a model file can execute code the moment it is loaded, and a fine-tuned open-weight model can carry a hidden backdoor. The course is built around serialisation formats (pickle, safetensors, GGUF), typosquatting on model hubs, poisoned datasets, dependency chains and licensing risks.

Hands-on work covers model scanning with ModelScan, model signing and verification, an approved model catalogue and building an AI-BOM with CycloneDX; participants prepare a model acceptance checklist for their organisation.

Who Should Attend?

  • Data scientists and ML engineers working with open-source and open-weight models
  • Platform and MLOps teams
  • Security teams responsible for software supply chain security
  • Technical leads involved in model procurement and licensing decisions

Key Takeaways

  • Explain the security differences between model file formats
  • Scan third-party models with tools such as ModelScan
  • Record model and dataset provenance
  • Produce an AI-BOM in CycloneDX format
  • Create a model acceptance and approval checklist for the organisation

Curriculum

Mapping the AI Supply Chain
  • Models, datasets, libraries and service providers
  • Supply chain risks in OWASP and MITRE ATLAS
Model File Risks
  • Code execution through pickle deserialisation
  • Comparing safetensors, GGUF and ONNX formats
Backdoored and Tampered Models
  • Triggers introduced via fine-tuning
  • Typosquatting and fake model cards on hubs
Scanning, Signing and Verification
  • Scanning with ModelScan and CI integration
  • Model signing, hash verification and an approved model catalogue
AI-BOM, Licensing and Provenance
  • Building an AI-BOM with CycloneDX
  • Assessing model and dataset licences

Other Courses in This Category

  • AI Agent Authentication, Authorization and Identity Security

    Teaches how to prevent confused deputy and privilege escalation risks in autonomous agents through workload identity, OAuth 2.0 On-Behalf-Of delegation, dynamic scope attenuation and OPA/Cedar policy enforcement.

    AI Security & Red Teaming Corporate
    Live Course
    5 modules
    View
  • AI Threat Modeling for LLM, RAG and Agent Systems

    Combines STRIDE with MITRE ATLAS and the OWASP Top 10 for LLM and Agentic AI to produce end-to-end AI threat models, trust boundary diagrams and control matrices for LLM, RAG and agent architectures.

    AI Security & Red Teaming Corporate
    Live Course
    5 modules
    View
  • AI Red Teaming for LLM, RAG and Agent Applications

    Teaches how to plan and execute scenario-driven red team campaigns against LLM, RAG and agent systems using PyRIT, garak and promptfoo, build multi-turn exploit chains and deliver reproducible findings.

    AI Security & Red Teaming Corporate
    Live Course
    5 modules
    View
  • LLM Guardrails and Multi-Layer Defense Design

    Teaches how to design and operate low-latency guardrails across input, retrieval, tool calling and output stages using NeMo Guardrails, Guardrails AI, Llama Guard and deterministic policy engines.

    AI Security & Red Teaming Corporate
    Live Course
    5 modules
    View
  • Automated AI Security Testing and CI/CD Security Evaluations

    Teaches how to build automated security evaluation suites for LLM and agent applications using promptfoo, garak and Inspect AI, enforce regression gates in CI/CD pipelines and continuously test model and prompt changes.

    AI Security & Red Teaming Corporate
    Live Course
    5 modules
    View
  • Secure Enterprise AI Reference Architecture

    Teaches how to design enterprise LLM, RAG and agent platforms on Zero Trust principles using a central AI gateway, network and tenant isolation, KMS key management, secure MCP integration and auditable telemetry.

    AI Security & Red Teaming Corporate
    Live Course
    5 modules
    View

Get in touch