The dataset includes selfies from over 1,000 people. Later, a team of over 200 people made 5,000+ replay display attacks based on these selfies. The attacks provide diversity of lighting, devices, and screens.