An attacker with access to a company's AI model API submits thousands of carefully chosen queries to reconstruct the model's decision boundary and build a functional replica. Which attack type is this?
Select an answer to reveal the explanation.
Short Explanation and Infographic
Here's the deal — b is correct because model extraction (also called model stealing) involves querying a target model's API to reconstruct its behavior and create a surrogate model — enabling intellectual property theft and enabling future adversarial attacks without needing API access. A corrupts training data.
Full explanation below image
Full Explanation
B is correct because model extraction (also called model stealing) involves querying a target model's API to reconstruct its behavior and create a surrogate model — enabling intellectual property theft and enabling future adversarial attacks without needing API access. A corrupts training data. C determines training set membership. D reconstructs training data from gradients in federated learning.