Adversarial Attack

This is a deliberate attempt to fool an AI model, often by making small, imperceptible changes to input data. These tiny modifications can cause the model to misclassify an image or misunderstand a command, even though the altered input looks identical to a human.