In this challenge, we release a large-scale surveillance FAS dataset, SuHiFiMask, including 101 participants of different ages, 232 masks and 200 2D attacks. A total of 10,195 videos were collected by 7 mainstream cameras in 40 real scenes. To the best of our knowledge, this is the first dataset to extend FAS to real surveillance scenes rather than mimicking low-resolution images and surveillance environments. We define three protocols for SuHiFiMask to fully evaluate the performance in surveillance environments: Protocol 1-ID, Protocol 2-Mask, and Protocol 3-quality. However, this challenge is based on Protocol 3 which evaluates the robustness of the algorithm to image quality degradation. Variable quality and disturbances are factors that affect the stability of the algorithm. Therefore, the robustness of the algorithm to quality degradation is an important metric to be evaluated. For more information, please refer to the Codalab webpage of our challenge.