QuestionQ96

VerifiedTesting AI-Specific Quality Characteristics

Which ONE of the following statements is a correct adversarial example in the context of machine-learning systems working with image classifiers?

Explanation

Adding known adversarial inputs to training data, known as adversarial training, can improve robustness but does not provide a foolproof prevention against adversarial examples. Defenses can be overcome by stronger adaptive attacks, so retraining with augmented examples alone cannot fully prevent such attacks.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!