Adaptive  1 Steganography in Extended Noisy Region

Abstract

A novel adaptive steganographic scheme for spatial image is proposed. A noisy function is used to measure texture complexity of 2 × 2 pixel blocks, which keeps monotonic increasing after ±1 modifications. Therefore, the message is embedded into the noisiest areas and the recipient can identify the embedding region. The 'double-layered embedding' is exploited to reduce the number of ±1 modifications, in which the fast matrix embedding and wet paper codes are applied to the least significant bit (LSB) plane and the second LSB plane, respectively. The experiments on resisting three steganalyzers show that the proposed method performs better than four typical steganographic schemes. Moreover, comparing with the extended highly undetectable steGO having parameter T = 255, the novel method achieves the competitive ability of resisting detection and faster embedding speed.

Parallel modular steganography using error images

“…The revised method simplifies DPMSE to facilitate understanding of the proposed parallel method. In this work we have the following: It is confirmed that the revised version is still less detectable compared to EA , HUGO , extended HUGO , and Adaptive ±1 Steganography . It is demonstrated that the proposed method runs faster than extended HUGO in serial. The revised version is re‐written, implementing parallel processing, and the effectiveness of the parallel code is calculated.…”

“…Regarding time performance, MS‐E could have executed faster if it had been implemented in C. It executed in MATLAB, which is slower than C code, a lower‐level language that generates relatively optimized machine code when compared to high‐level languages such as MATLAB. In the next section, we will prove that, unlike that has block dependence in noisy regions, the proposed method can be executed more quickly because it has the ability to be parallelized. The current method is a semi‐blind steganography method.…”

“…The execution time of the proposed algorithm is approximately 2.32 times faster than the extended HUGO method in serial mode. Adaptive ±1 Steganography in Extended Noisy Regions is reported to be 3.93 times faster than the extended HUGO using a C implementation. However, it can be detected more easily than MS‐E (Table ).…”

A further study of large payloads matrix embedding

“…For example, for q/p = 1/2, the best parameters are (a, b, m, k 1 ) = (20, 37,6,18) where the dimensions of H and B are 120 × 240 and 20 × 37, respectively. On the other hand, reviewing (41), if k 1 = c, it means that all the vectors in F k 1 2 are tested for minimizing (36) and a coset leader is thus selected without any loss of optimality.…”

“…In this light, a common viewpoint is that the content adaptive approach for steganography has the potential to provide a higher level of security. Up to now, many content adaptive steganographic methods are already proposed [7,8,11,[18][19][20][21][27][28][29]33,39,40,42]. ME is an effective steganographic scheme which can improve the embedding efficiency and enhance the stego-security [1,12,13,23,26,31].…”