Program 4 · Data Protection, Privacy Engineering and AI Security

Status: not started

Data protection, privacy engineering, and the new attack surface introduced by machine learning systems. Expected topics include:

Data & privacy

Data classification and handling
Privacy-by-design and privacy engineering
GDPR, CCPA, and other major privacy regimes
Data minimization, purpose limitation, retention
Anonymization, pseudonymization, differential privacy

AI security

The AI attack surface: training data, models, prompts, outputs
Data poisoning and backdoor attacks
Model extraction and model inversion
Adversarial examples
Prompt injection and jailbreaking for LLMs
Membership inference and privacy leakage
Securing the ML pipeline end-to-end

Modules and submodules will be added as the program is started.

Feeds from

Program 1 — confidentiality and compliance framing.
Program 3 — threat modeling techniques adapted for ML systems.

Feeds into

Program 5 — security is a prerequisite for responsibility; a system that cannot be trusted to keep data safe cannot be responsible.