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 14 hours
Course Outline
The AI Threat Landscape
- Why AI security stands apart: addressing non-determinism, opaque reasoning, and prompts as attack surfaces
- Categorizing attacks: training-time vs. inference-time vs. supply chain threats
- Defining the ML adversary model: understanding who targets AI systems and their motivations
OWASP Top 10 for LLM Applications
- Prompt injection: analyzing direct and indirect attack vectors
- Managing insecure output handling and cross-plugin request forgery
- Addressing training data poisoning and supply chain vulnerabilities
- Mitigating model denial of service, sensitive information disclosure, and excessive agency
- Hands-on lab: actively exploiting each OWASP category against a test application
Prompt Injection and Jailbreak Red Teaming
- Classifying injection techniques: direct, indirect, multi-turn, and multi-modal approaches
- Automating red-teaming using tools like Giskard, Garak, and custom fuzzing frameworks
- Evaluating jailbreak classification methods and defense effectiveness
- Creating a red-team harness for sustained LLM security testing
Model-Level Attacks and Defenses
- Model extraction: preventing the theft of model weights and functionality via API queries
- Membership inference: protecting against determining if specific data was part of the training set
- Adversarial examples: countering perturbations designed to deceive classifiers and embeddings
- Data poisoning: mitigating corruption of training data that could induce backdoors or degrade performance
Input and Output Security Controls
- Input sanitization strategies that go beyond traditional web security defenses
- Output filtering: addressing toxicity, PII leakage, and the risk of hallucinated code execution
- Implementing guardrails as security infrastructure using NeMo, Guardrails AI, and custom policies
- Enforcing structured outputs to establish a robust security boundary
AI Supply Chain Security
- Model provenance: verifying the authenticity and integrity of models
- Scanning dependencies within ML frameworks and model formats
- Securing model serving through sandboxing, network isolation, and least-privilege access
- Vetting fine-tuned and community models for embedded malware
Operational Security for AI Systems
- Implementing access control for model endpoints, vector stores, and agent tools
- Maintaining comprehensive audit logs for every model interaction and decision
- Responding to incidents involving AI-specific breaches, including when the model itself is compromised
- Integrating continuous security testing into CI/CD pipelines for ML workflows
Building an AI Security Program
- Developing an AI security maturity model and strategic roadmap
- Integrating AI security into existing AppSec and cloud security initiatives
- Navigating governance frameworks and emerging regulations for AI systems
- Creating and sustaining an organizational AI security playbook
Requirements
- Practical experience in deploying ML models or LLM applications in production environments.
- A solid understanding of security fundamentals, including authentication, authorization, and threat modeling.
- Proficiency in Python for executing adversarial testing exercises.
Target Audience
- Security engineers looking to expand their expertise into AI/ML threat landscapes.
- ML engineers accountable for model safety and system robustness.
- Red team members incorporating AI systems into their security testing scope.
Testimonials (1)
inventory and identifying the different risk exposures within AI
