Steganography Performance over AWGN Channel

Abstract

Steganography can be performed using frequency domain or spatial domain. In spatial domain method, the least significant bits (LSB) is the mostly used method where the least significant bits of the image's pixels binary representation are used to carry the confidential data bits. On the other hand, secret data bits in the frequency domain technique are hidden using coefficients of the image frequency representation such as discrete cosine transform (DCT). Robustness against image attacks or channel's noise is a key requirement in steganography. In this paper, we study the performance of the steganography methods over a channel with Added White Gaussian Noise (AWGN). We use the bit error rate to evaluate the performance of each method over a channel with different noise levels. Simulation results show that the frequency domain technique is more robust and achieves better bit error rate in a noisy channel than the spatial domain method. Moreover, we enhanced the steganography system robustness by using convolution encoder and Viterbi decoder. The effect of the encoder’s parameters, such as rate and constraint length is evaluated.