Dynamic energy based text-in-speech spectrum hiding using speech masking properties

In this paper we present a new method for text-in-speech Fast Fourier Transform (FFT) based steganography. Our method exploits high energetic magnitude frequency components of speech spectrum to hide the secret text. The energy level of each magnitude component is evaluated, and considered only if it is high enough to embed secret data. A frequency masking threshold is then calculated for each energetic frequency magnitude component to decide where to embed the secret text. As a result, the initial energy of the cover speech spectrum is slightly modified and the distortion error is always kept under the mask. The objective and subjective evaluations show that introducing the energy-based frequency masking opens new frontiers in text-in-speech steganography since higher bit rate text embedding was achieved while obtaining a stego-speech with unnoticeable distortion.