Adversarial attacks on AI

Adversarial attacks craft inputs that fool AI — imperceptible image changes cause misclassification. Understanding attacks helps build robust defenses.

• Adversarial inputs deliberately fool AI systems.
• Understanding attacks improves defenses.