Model Inversion Attack
An attack that attempts to reconstruct training data or sensitive features of individuals by exploiting access to a trained machine learning model's predictions.
In plain language
An attack where someone uses an AI's outputs to reverse-engineer private training data. Like using a facial recognition system to reconstruct actual faces from its training set.
Why this matters
Model inversion directly threatens data privacy. Your AI risk assessment process must evaluate this threat for models trained on sensitive personal data and implement appropriate safeguards.
Relevance
GovernanceModel inversion risk assessment and mitigation are essential components of privacy governance for any AI trained on personal or sensitive data.
Related terms
Putting model inversion attack into practice in your organisation?
Ready to transform your AI strategy?
Partner with Australia's AI strategy and governance specialists. From adoption roadmaps to ISO 42001 audit readiness.