A Real-time Implementation for the Speech Steganography using Short-Time Fourier Transformior Secured Mobile Communication

Abstract

Steganography can be described as approach of masking an undisclosed message with a normal message which is known as the Carrier message signal. DSP techniques, such as LSB encoding, have historically been implemented for secret information hiding. Utilization ofsteganography functions of deep neural networks for voice data is something this paper will present. This paper also demonstrate that the steganography techniques suggested for vision are less suitable for speech signals this paper present a implementation technique that involves the use of ISTFT and STFT as differentiablelayers in the network. Empirically, the efficacy of the proposed methods based on multiple datasets of speech should be demonstrated and the outcome are examined quantitatively and qualitatively. Using of multiple decoders or a single conditional decoder helps to hide multiple signals in a single carrier signal. Finally, under various channel distortion situations, this model Qualitative studies indicate that human listeners cannot detect changes made to the carrier and hence the decoded messages are highly intelligible.