Omed Khalind and Benjamin Aziz, University of Portsmouth, United Kingdom

In this paper, we propose a new method of non-adaptive LSB steganography in still images to improve the embedding efficiency from 2 to 8/3 random bits per one embedding change even for the embedding rate of 1 bit per pixel. The method takes 2-bits of the secret message at a time and compares them to the LSBs of the two chosen pixel values for embedding, it always assumes a single mismatch between the two and uses the second LSB of the first pixel value to hold the index of the mismatch. It is shown that the proposed method outperforms the security of LSB replacement, LSB matching, and LSB matching revisited by reducing the probability of detection with their current targeted steganalysis methods. Other advantages of the proposed method are reducing the overall bit-level changes to the cover image for the same amount of embedded data and avoiding complex calculations. Finally, the new method results in little additional distortion in the stego image, which could be tolerated.

Steganography, Embedding efficiency, Probability of detection, Single Mismatch, LSB matching, LSB replacement