感謝您提交詢問！我們的一位團隊成員將在短時間內與您聯繫。
感謝您提交預訂！我們的一位團隊成員將在短時間內與您聯繫。
課程簡介
The AI Threat Landscape
- Why AI security is different: non-determinism, opaque reasoning, prompt as attack surface
- Attack taxonomy: training-time vs inference-time vs supply chain attacks
- The ML adversary model: who attacks AI systems and why
OWASP Top 10 for LLM Applications
- Prompt injection: direct and indirect attack vectors
- Insecure output handling and cross-plugin request forgery
- Training data poisoning and supply chain vulnerabilities
- Model denial of service, sensitive information disclosure, and excessive agency
- Hands-on lab: exploiting each OWASP category against a test application
Prompt Injection and Jailbreak Red Teaming
- Taxonomy of injection techniques: direct, indirect, multi-turn, and multi-modal
- Automated red-teaming with Giskard, Garak, and custom fuzzing tools
- Jailbreak classification and defense evaluation
- Building a red-team harness for continuous LLM security testing
Model-Level Attacks and Defenses
- Model extraction: stealing model weights and functionality via API queries
- Membership inference: determining if data was in the training set
- Adversarial examples: perturbations that fool classifiers and embeddings
- Data poisoning: corrupting training data to induce backdoors or degrade performance
Input and Output Security Controls
- Input sanitization beyond traditional web defenses
- Output filtering: toxicity, PII leakage, hallucinated code execution
- Guardrails as security infrastructure: NeMo, Guardrails AI, and custom policies
- Structured output enforcement as a security boundary
AI Supply Chain Security
- Model provenance: verifying model authenticity and integrity
- Dependency scanning for ML frameworks and model formats
- Secure model serving: sandboxing, network isolation, and least-privilege access
- Vetting fine-tuned and community models for embedded malware
Operational Security for AI Systems
- Access control for model endpoints, vector stores, and agent tools
- Audit logging for every model interaction and decision
- Incident response for AI-specific breaches: when the model itself is compromised
- Continuous security testing in CI/CD for ML pipelines
Building an AI Security Program
- AI security maturity model and roadmap
- Integrating AI security into existing AppSec and cloud security programs
- Governance frameworks and emerging regulations for AI systems
- Creating and maintaining an organizational AI security playbook
最低要求
- Experience deploying ML models or LLM applications in production.
- Familiarity with security concepts including authentication, authorization, and threat modeling.
- Python proficiency for adversarial testing exercises.
Audience
- Security engineers expanding into AI/ML threat surfaces.
- ML engineers responsible for model safety and robustness.
- Red team members adding AI systems to their testing scope.
14 小時
