Ciphertext only attack on QR code optical encryption system with spatially incoherent illumination using a neural network

Abstract Optical encryption methods attract a lot of attention owing to their high encryption speed and bandwidth. Recently, neural networks (NNs) have been used for cryptanalysis of optical encryption techniques. In this paper, we for the first time to our knowledge applied a NN for ciphertext only attack on an optical encryption system with spatially incoherent illumination. A NN was used to extract encryption keys from ciphertexts, which can be used to decrypt the plaintext QR codes. Additionally, an optically encrypted QR code was successfully decoded after using the key extracted by the trained NN, that has been processed to account for discrepancies between the numerical model and the optical setup. The results show the vulnerability of the existing optical encryption system with incoherent light to attacks of this type, which indicates the need for improved optical encryption security.