Deep Learning Based Attack on Phase-Truncated Optical Encoding

Abstract

We apply the learning based attack to study the vulnerability of phase-truncated optical encoding scheme. By using a number of ciphertext-plaintext pairs to train a designed learning model, an attacker can effectively analyze the vulnerability of optical encryption scheme based on phase truncation. The learning based attacks for phase-truncated optical encoding can retrieve unknown plaintexts from the given ciphertexts, which can avoid the retrieval of security keys and the design of complex phase retrieval algorithms. It is demonstrated that the learning based attack can provide a promising approach for vulnerability analysis of phase-truncated optical cryptosystems.