Cryptanalysis Of Phase Information Based On A Double Random-Phase Encryption Method

W Song,X Liao,D Weng,Y Zheng,Y Liu,Y Wang
DOI: https://doi.org/10.1016/j.optcom.2021.127172
IF: 2.4
2021-01-01
Optics Communications
Abstract:Phase-only information derived using a double random-phase encryption (DRPE) algorithm is not visually recognizable but can be authenticated using a nonlinear cross-correlation method. This can be considered to enhance DRPE security (which is vulnerable to a chosen-plaintext attack). Here, we show that phase-only images derived using the DRPE method remains vulnerable to such an attack; the at-risk images include full and sparse-phase images and phase information with noise. We develop an encoder-decoder deep-learning method to decrypt phase-only images encrypted using the DRPE method. The deep-learning structure can be trained using only spare phase information from the encrypted image; experimentally, the trained model readily decrypted phase-only or partial-phase-encrypted images based on the DRPE algorithm.
What problem does this paper attempt to address?