Desighning an Image Steganography Algorithm Based on Entropy and ELSB2

Security, robustness, and capacity are the most important evaluation criteria in steganography. The aim of security is reducing the probability of message detection in an image and the aim of robustness is stabilizing the embedded message against removal and distortion due to intentional and unintentional attacks, such as noise and filtering. In this paper, an efficient adaptive image steganographic algorithm is designed in spatial domain which is also usable in frequency domain. In this algorithm, roughness (pixel changes) of each non-overlapping 3×3 window in the image is calculated using entropy of corresponding weighted graph. Each window threshold roughness is calculated based on length of message and overall image texture. Hiding message is done with specified density in LSB2 of 4 pixels of rough windows of an image by considering xor operator and randomization. Simulation on 5000 natural images and applying a modern steganalytic algorithm, SRM, shows that in comparison with other popular novel method, the proposed method has increased the security (Reducing the probability of message detection) about 2% in addition to robustness improvement.