Cryptanalysis of phase information based on a double random-phase encryption method

Phase-only information derived using a double random-phase encryption (DRPE) algorithm is not visually recognizable but can be authenticated nonlinear cross-correlation method. This considered to enhance DRPE security (which vulnerable chosen-plaintext attack). Here, we show that phase-only images the method remains such an attack; at-risk include full- and sparse-phase phase with noise. We develop encoder–decoder deep-learning decrypt encrypted The structure trained only spare from image; experimentally, model readily decrypted or partial-phase-encrypted based on algorithm. • have developed simulate chosen-ciphertext attack created by DRPE. robustly extracted plaintext when were sheared polluted additive multiplicative Image authentication via was possible. shown