Abstract: Modern face recognition systems widely use deep convolutional neural networks (DCNNs). However, DCNNs are susceptible to adversarial examples, posing security risks to these systems.