E1 · ML Lifecycle specialization
ML Lifecycle and Adversarial ML Security
Assess and secure data and model pipelines and test representative training-time and inference-time attacks.
Proposed
0 of 6 modules accessible. Joining Skool does not unlock planned modules.
At a glance
- Audience
- Security practitioners taking responsibility for AI systems.
- Prerequisites
- Core pathway or equivalent
- Basic statistics
- Supervised learning
- Train, validation, and test separation
- A simple ML framework
- Planned duration
- 36 hours. Assumes the entry requirements are met. Planning estimate, not a measured completion time. Per module: 2 to 4 hours, plus practice.
- Assessment
- Reproducible pipeline experiments with rubric.
- What is free
- The free foundation module listed on each course page, as it releases, plus community access.
Job tasks this course prepares you for
- Trace dataset and model lineage through the supplied pipeline.
- Detect and investigate a seeded unauthorized dataset or model change.
- Reproduce one poisoning attack and one inference-time attack in a controlled small-model exercise.
- Measure a mitigation's effect on both attack success and legitimate model performance.
- Implement a promotion policy that ties the approved artifact to its evaluation evidence.
Learning outcomes
- Trace dataset and model lineage through the supplied pipeline.
- Detect and investigate a seeded unauthorized dataset or model change.
- Reproduce one poisoning attack and one inference-time attack in a controlled small-model exercise.
- Measure a mitigation's effect on both attack success and legitimate model performance.
- Implement a promotion policy that ties the approved artifact to its evaluation evidence.
Modules
00Lifecycle and attacker accessProposed
Topics
- datasets, labels, features, training jobs, registry, serving.
01Dataset securityProposed
Topics
- lineage, entitlement, integrity, leakage, contamination, reproducibility.
02Poisoning and backdoorsProposed
Topics
- controlled examples, limits of validation, remediation tradeoffs.
03Inference-time attacksProposed
Topics
- evasion, model extraction, membership inference, model inversion concepts.
04Pipeline controlsProposed
Topics
- isolated training, secrets, experiment integrity, artifact promotion, rollback.
05Evaluation and residual riskProposed
Topics
- utility and security tradeoffs, attacker assumptions, unsupported defenses.
Labs and deliverables
- Secure a small predictive-model pipeline. Lineage records, an attack and defense notebook, pipeline changes, and a limitations report.
Labs run on SIFT, the forensics workstation Rob T. Lee created, on your own machine. Each module states what it needs before you start. Each module confirms its setup requirements and any cloud or API costs before it opens. About the SIFT lab environment.
Assessment
Reproducible pipeline experiments with rubric. A named assessor grades against the published rubric (rubric to be published before release).
Pass criteria
- Reproducible experiments
- Enforced promotion controls
- Correct attacker assumptions
- 80% rubric
Critical failures
- Unauthorized cross-tenant access
- Out-of-scope activity
- Uncontrolled destructive action
- Fabricated evidence
- Exposed credentials
What this course does not teach
- Not comprehensive adversarial ML research, frontier-model training, or formal privacy expertise; does not prove models are free of backdoors.