This paper proposes a chaotic fingerprint images encryption approach. An image of a fingerprint is encrypted via a two-dimensional (2D) chaotic sequence obtained from multi-scroll chaotic attractors. Initial values of the chaotic attractors are served as the private key, which can be generated from the pixel distribution of the binary images of the captured fingerprints. Due to the dynamic uncertainties in the acquisition process of fingerprint images, the keys generated from the pixel value distribution are virtually random. With the elaborately designed 2D chaotic sequence, the encrypted fingerprint images have balanced 0–1 ratio and ideal nonlinearity. Only with the valid private key can the images of fingerprint be recovered. Simulation results and 2D-DFT validate this chaotic encryption approach.