A multi-threshold based audio steganography scheme

Abstract

Human auditory system is highly sensitive that can detect or discriminate subtle changes in an audio. The direct least significant bit (LSB) substitution scheme, often used in image steganography, cannot be effective for audio steganography to provide secrecy needed for many audio applications, because the embedded secret bits in the LSBs of the audio samples in a silent or near silent segment of an audio can produce strong hissing noise when played. In this work, governed by an error criterion, we present a multi-threshold-based scheme for audio steganography. In order to evade detection by means of any statistical or perceptual analysis of the resultant audio, the threshold values are derived from an error criterion that can be used to limit the number of bits to be embedded in an audio sample. To defend against retrieval of bits by an adversary as well as to dissipate any pattern in the resultant audio, a scheme for chaotic distribution of the secret message bits in the cover audio is also proposed. In comparison with most recently proposed schemes in literature, our proposed scheme exhibits better performance against statistical and perceptual analysis for the same embedding capacity. Our proposed scheme can be used to increase the capacity of the cover audio in a controlled manner by adjusting the error criterion.