Deliberate attempts to fool machine learning models by introducing carefully crafted perturbations to input data that cause misclassification or incorrect outputs. Adversarial attacks expose model vulnerabilities and highlight the need for robust AI systems in security-critical applications.
