Nonbinary audio cryptography

Abstract

Visual cryptography, introduced by Naor-Shamir at Eurocrypt’94, only requires primitive technology to decrypt the ciphertext. However, a disadvantage of it is that the “ciphertext”, as a random looking transparency, is suspicious to a censor. The solutions recently proposed by Desmedt-Hou-Quisquater to avoid these problems are neither user friendly, having a low bandwidth, nor are tested. In this paper we present three schemes that overcome these problems. As in one of the Desmedt-Hou-Quisquater’s schemes, a share (or a ciphertext) corresponds to an audio signal, such as music. While in the Desmedt-Hou- Quisquater scheme the plaintext was binary, in our schemes the plaintext can also be speech, or any other audio signal. By introducing variations of the one-time pad we guarantee perfect secrecy. The ciphertext is nonsuspicious, when tested with human ears, is indistinguishable from normal music.