Hacky Certified AI Security Specialist
HCAIS · Level 2Introduction
Protect machine-learning models and AI systems. Three levels, each self-paced, each with its own exam and certificate.
Buy now- Author: Gaurav Sharma
- Level: Level 2 · Practitioner
- Study time: 48 hours
- Certification: Hacky Certified AI Security Specialist
Curriculum
6 modules · 24 hours of guided learning · self-paced, with hands-on labs
Job-oriented, scenario-led progression in hacky certified ai security specialist. Exercises are conducted only in authorized training environments. The level builds on previous levels while adding distinct techniques, deeper evidence requirements and more complex capstones.
Prerequisites: Basic IT literacy and willingness to use Linux/Windows command lines.
Tool stack: Python, Ollama, PyTorch, scikit-learn, Garak, PyRIT, Promptfoo, MLflow, model scanners
6h guided instruction + 13h guided/independent practical labs + 5h reading, review and assessment. These are total learning hours, not 24 hours of live lectures.
Module 01AI/ML lifecycle and threat modeling4 h
Coverage. Identify core concepts, demonstrate the workflow, and record reliable evidence.
Practical. Complete a guided lab with baseline and expected evidence.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 02ML data provenance and governance4 h
Coverage. Identify core concepts, demonstrate the workflow, and record reliable evidence.
Practical. Complete a guided lab with baseline and expected evidence.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 03OWASP GenAI and AI security risk frameworks4 h
Coverage. Identify core concepts, demonstrate the workflow, and record reliable evidence.
Practical. Complete a guided lab with baseline and expected evidence.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 04Model interfaces, APIs and access controls4 h
Coverage. Identify core concepts, demonstrate the workflow, and record reliable evidence.
Practical. Complete a guided lab with baseline and expected evidence.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 05Dataset poisoning and integrity testing4 h
Coverage. Identify core concepts, demonstrate the workflow, and record reliable evidence.
Practical. Complete a guided lab with baseline and expected evidence.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 06Adversarial examples and robustness4 h
Coverage. Identify core concepts, demonstrate the workflow, and record reliable evidence.
Practical. Complete a guided lab with baseline and expected evidence.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Guided scenario: demonstrate correct methodology, collect evidence and submit a short structured report.
20% knowledge checks, 60% hands-on practical, 20% evidence-based reporting Recommended pass: 70% overall with mandatory completion of safety and evidence checkpoints.
Final exam
A 2-hour online exam with a 70% pass mark. Recommended pass: 70% overall with mandatory completion of safety and evidence checkpoints. Attempts are free until you pass. The exam opens 7 days after purchase.
Curriculum
9 modules · 48 hours of guided learning · self-paced, with hands-on labs
Job-oriented, scenario-led progression in hacky certified ai security specialist. Exercises are conducted only in authorized training environments. The level builds on previous levels while adding distinct techniques, deeper evidence requirements and more complex capstones.
Prerequisites: Completion of Level 1 or demonstrably equivalent practical skills.
Tool stack: Python, Ollama, PyTorch, scikit-learn, Garak, PyRIT, Promptfoo, MLflow, model scanners
12h guided instruction + 26h guided/independent practical labs + 10h reading, review and assessment. These are total learning hours, not 48 hours of live lectures.
Module 01Model interfaces, APIs and access controls6 h
Coverage. Investigate a realistic scenario, compare alternative approaches, tune decisions and document remediation.
Practical. Complete a case-based lab using multiple evidence sources and a verification checklist.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 02Dataset poisoning and integrity testing6 h
Coverage. Investigate a realistic scenario, compare alternative approaches, tune decisions and document remediation.
Practical. Complete a case-based lab using multiple evidence sources and a verification checklist.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 03Adversarial examples and robustness6 h
Coverage. Investigate a realistic scenario, compare alternative approaches, tune decisions and document remediation.
Practical. Complete a case-based lab using multiple evidence sources and a verification checklist.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 04Model extraction and privacy risk5 h
Coverage. Investigate a realistic scenario, compare alternative approaches, tune decisions and document remediation.
Practical. Complete a case-based lab using multiple evidence sources and a verification checklist.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 05Prompt injection in AI-enabled applications5 h
Coverage. Investigate a realistic scenario, compare alternative approaches, tune decisions and document remediation.
Practical. Complete a case-based lab using multiple evidence sources and a verification checklist.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 06AI supply chain and dependency assessment5 h
Coverage. Investigate a realistic scenario, compare alternative approaches, tune decisions and document remediation.
Practical. Complete a case-based lab using multiple evidence sources and a verification checklist.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 07Secure model deployment and monitoring5 h
Coverage. Investigate a realistic scenario, compare alternative approaches, tune decisions and document remediation.
Practical. Complete a case-based lab using multiple evidence sources and a verification checklist.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 08LLM evaluation and red teaming5 h
Coverage. Investigate a realistic scenario, compare alternative approaches, tune decisions and document remediation.
Practical. Complete a case-based lab using multiple evidence sources and a verification checklist.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 09Agent tools, permissions and sandboxing5 h
Coverage. Investigate a realistic scenario, compare alternative approaches, tune decisions and document remediation.
Practical. Complete a case-based lab using multiple evidence sources and a verification checklist.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Multi-system case: independently correlate artifacts, prioritize findings and propose tested remediation.
20% knowledge checks, 60% hands-on practical, 20% evidence-based reporting Recommended pass: 70% overall with mandatory completion of safety and evidence checkpoints.
Final exam
A 24-hour practical exam in a hosted lab, then 24 hours to submit your report. Recommended pass: 70% overall with mandatory completion of safety and evidence checkpoints. Attempts are free until you pass. The exam opens 7 days after purchase.
Curriculum
9 modules · 72 hours of guided learning · self-paced, with hands-on labs
Job-oriented, scenario-led progression in hacky certified ai security specialist. Exercises are conducted only in authorized training environments. The level builds on previous levels while adding distinct techniques, deeper evidence requirements and more complex capstones.
Prerequisites: Completion of Level 2 or equivalent professional experience; comfortable with independent investigations.
Tool stack: Python, Ollama, PyTorch, scikit-learn, Garak, PyRIT, Promptfoo, MLflow, model scanners
18h guided instruction + 40h guided/independent practical labs + 14h reading, review and assessment. These are total learning hours, not 72 hours of live lectures.
Module 01Prompt injection in AI-enabled applications8 h
Coverage. Lead an integrated assessment, validate assumptions, automate repeatable checks and produce defensible findings.
Practical. Complete a scenario-driven challenge, defend findings, and validate controls or mitigations.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 02AI supply chain and dependency assessment8 h
Coverage. Lead an integrated assessment, validate assumptions, automate repeatable checks and produce defensible findings.
Practical. Complete a scenario-driven challenge, defend findings, and validate controls or mitigations.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 03Secure model deployment and monitoring8 h
Coverage. Lead an integrated assessment, validate assumptions, automate repeatable checks and produce defensible findings.
Practical. Complete a scenario-driven challenge, defend findings, and validate controls or mitigations.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 04LLM evaluation and red teaming8 h
Coverage. Lead an integrated assessment, validate assumptions, automate repeatable checks and produce defensible findings.
Practical. Complete a scenario-driven challenge, defend findings, and validate controls or mitigations.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 05Agent tools, permissions and sandboxing8 h
Coverage. Lead an integrated assessment, validate assumptions, automate repeatable checks and produce defensible findings.
Practical. Complete a scenario-driven challenge, defend findings, and validate controls or mitigations.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 06RAG trust boundaries and retrieval security8 h
Coverage. Lead an integrated assessment, validate assumptions, automate repeatable checks and produce defensible findings.
Practical. Complete a scenario-driven challenge, defend findings, and validate controls or mitigations.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 07AI logging, telemetry and incident response8 h
Coverage. Lead an integrated assessment, validate assumptions, automate repeatable checks and produce defensible findings.
Practical. Complete a scenario-driven challenge, defend findings, and validate controls or mitigations.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 08AI security assessment methodology8 h
Coverage. Lead an integrated assessment, validate assumptions, automate repeatable checks and produce defensible findings.
Practical. Complete a scenario-driven challenge, defend findings, and validate controls or mitigations.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Module 09AI security capstone8 h
Coverage. Lead an integrated assessment, validate assumptions, automate repeatable checks and produce defensible findings.
Practical. Complete a scenario-driven challenge, defend findings, and validate controls or mitigations.
Evidence. annotated observations, reproducible steps, risk/impact notes and a concise technical recommendation.
Professional engagement: plan, execute and defend a complex scenario with peer review and executive/technical reporting. HACKY SECURITY | CERTIFICATION CURRICULUM | 2026 3
20% knowledge checks, 60% hands-on practical, 20% evidence-based reporting Recommended pass: 70% overall with mandatory completion of safety and evidence checkpoints.
Final exam
A 48-hour scenario exam, assessed on your written report. Recommended pass: 70% overall with mandatory completion of safety and evidence checkpoints. Attempts are free until you pass. The exam opens 7 days after purchase.
What you receive
Course author
Gaurav Sharma
Founder & CISO, Hacky SecurityOffensive and Defensive Security Expert
Gaurav founded Hacky Security and leads its SOC, curriculum and AI security work. He has trained more than 40,000 professionals and has taught AI security at IIT Madras.