Question 30
Domain 2 — AI Operations, Lifecycle, and Control EnvironmentWhy are AI systems particularly vulnerable to model inversion attacks?
Correct answer: B
Explanation
AI systems can leak information because their outputs may reveal patterns tied to the training data. In a model inversion attack, an attacker uses those outputs to “infer sensitive input data from model outputs,” exposing private details that the model has learned.
Why each option is right or wrong
A. AI models are immune to external probing
B. Attackers can infer sensitive input data from model outputs
Model inversion exploits the fact that a trained model’s outputs can expose information about the data it was trained on, allowing an attacker to reconstruct or infer private inputs from prediction scores, probabilities, or labels. The vulnerability is greatest where outputs are detailed or confidence-bearing, because those signals can be used to reverse-engineer sensitive attributes from the model’s behavior.
C. Inversion attacks encrypt datasets to hide sensitive variables
D. They only target traditional software, not AI