Adversarial example
An adversarial example is a specially crafted input designed to deceive machine learning models, causing them to make incorrect predictions or classifications. These examples exploit vulnerabilities in models, often with minimal perturbations imperceptible to humans.