Abstract: Convolutional neural networks (CNNs) are known to be vulnerable to adversarial examples even in black-box scenarios, posing a significant threat to their reliability and security. Most ...