A new hybrid steganographic method for histogram preservation

A new hybrid steganographic method for histogram

preservation

Umesh Ghanekar

Abstract This paper presents a histogram preserving data

embedding method for grey-scale images which is based on pixel

value differencing (PVD) and least-significant-bit (LSB)

substitution methods. Various PVD based steganographic

methods achieve high data embedding capacity with minimum

distortions in stego image at the cost of change in histogram

characteristics which is can be detected by histogram based

steganalysers. This persistent problem can been taken care off by

proposed method of data hiding. The improved performance of

the proposed method is verified through extensive simulations.

Keywordssteganography; PVD; embedding capacity;

histogram characteristics;

I. INTRODUCTION

In recent years, steganography has emerged as an

interesting area of research. Steganography is basically used to

enhance the communications security. It hides the very

existence of the secret message into the cover media such as

digital image, audio, video, text etc [1]. In this paper, grey-

scale digital images have been used as the cover media for

hiding the secret message.

Many data hiding methods have been proposed so far and

among them the most simple and well- known steganography

method is least-significant-bit (LSB) replacement. Here, the

secret message is concealed directly into the LSBs of each

pixel of an image. This direct embedding procedure of various

existing spatial domain steganographic methods like LSB

replacement and others is incapable of exploiting the true

embedding capacity of any cover image. An image consists of

two areas i.e. edge area and smooth area. Edge areas can be

embedded with more number of bits than smooth areas, as

edges are less sensitive towards the changes in pixel

intensities. In 2003, Wu and Tsai used this concept and

presented a steganography method using PVD [2]. This

method hides different amount of secret bits in consecutive

non-overlapping pixel pairs by taking the difference value

between the pixels of a pixel pair. Further to increase the

embedding capacity a hybrid method based on PVD and fixed

sized LSB method was proposed by Wu et al. [3]. In 2008,

another hybrid method was presented based on PVD and

modulus function [4]. This method provides higher

imperceptibility of the stego image than the previous methods

while maintaining good data embedding capacity. An adaptive

LSB replacement method was also proposed in 2008 which

utilises the basic concept of data hiding based on human visual

system (HVS) [5]. As a result, pixels are embedded with

different number of secret bits using LSB replacement

method. In the year 2012, a novel adaptive data hiding method

based on LSB substitution and PVD was proposed [6]. This

method is able to conceal large amount of secret data and

provide good stego image quality but is unable to preserve

histogram characteristics.

Here, we have proposed a steganographic method using

LSB substitution and PVD in order to preserve the image

histogram. In this method, we have increased the block size to

3 3 as compared to 1 3 of adaptive LSB and PVD [6]. The central pixel of each block is termed as base pixel and 3-bits

are embedded in this pixel with the help of LSB replacement

method and optimal pixel adjustment process (OPAP) [7].

Remaining pixels of the block are embedded with secret data

bits using PVD. The performance of the proposed method is

demonstrated through extensive simulations.

The paper is organized as follows. Section II presents the

proposed method. Experimental results are shown in section

III. Finally, conclusions are given in section IV.

II. PROPOSED METHOD

This section deals with the procedure of proposed method which consists of three phases, namely, the range division phase, the embedding phase and the extracting phase. These phases are described as follows.

A. Range division phase

Prior to embedding the secret message, the grey level range [0,255] is divided into five ranges where , denotes the lower bound of the range and denotes the upper bound of the range . These five ranges can be , , , and . Fig. 1 shows the dividing case i.e. div=31 for the proposed method. It divides the range [0,255] into lower level which consist of ranges , , and higher level which include ranges , . Let are the number of bits to be embedded in the pixels falling under the range . According to HVS, changes in edge areas are less visible than smooth areas and hence more data can be

Priya darshni,

Dept. of Electronics and Communication Engineering National Institute of Technology

Kurukshetra, India

[email protected], [email protected]

NaveenTypewritten textInt. Journal of Electrical & Electronics Engg. Vol. 2, Spl. Issue 1 (2015)e-ISSN: 1694-2310 | p-ISSN: 1694-2426

NaveenTypewritten text139 NITTTR, Chandigarh EDIT-2015

embedded in edges. In the proposed method, first three ranges ( fall under the category of smooth regions whereas last two ranges falls in the edge regions. Therefore, we propose to embed bits in the lower level and bits in the higher level.

B. Embedding phase

The cover image is divided into consecutive non-

overlapping blocks of size 3 3 in raster scan manner. Each block has a centre pixel named as base pixel . Data embedding in each block is performed by the following steps

as given in [6].

Step 1: Consider 3-rightmost LSBs of and transform these three LSBs to a decimal value, say . Read 3-bits from binary secret data in continuation and replace the 3 LSBs of

with these binary secret data bits to obtain . Also,

transform these bits to a decimal value, say .

Step 2: Compute the difference value using .

Step 3: Modify using OPAP as follows

{

(1)

Step 4: Compute the absolute difference values between the

base pixel and other pixels of the block by using

| | (2)

where and denotes the location of the pixel in a block. Therefore, eight difference values are calculated.

Step 5: Assign the ranges corresponding to the differences found in Step 4 and obtain the lower bounds too i.e. .

Accordingly, calculate which denotes the number of bits to

be concealed into eight pixels.

Step 6: Read bits in continuation from the binary secret

message and transform these bit-sequences into decimal

values, say . Now, compute the new difference values

using

(3)

Step 7: Calculate the two new values of each pixel of a block

using

(4)

Step 8: Choose the best new value for these pixels from the

values obtained in Step 7 using

{

|

| |

|

(5)

Repeat the above procedure for every block of the cover image so as to obtain the final stego image.

C. Extracting phase

At first, the stego image is divided into consecutive non

overlapping blocks of size 3 3 and then for the complete extraction of the secret message following steps are executed.

Step 1: Select the centre pixel as the base pixel and extract 3-

LSB bits from it. Call this binary sequence as .

Step 2: Calculate the absolute difference values between

the base pixel and the other pixels of a block and then find the

range to which these difference values belong to. Then, obtain

the lower bound of the corresponding range and also

determine the number of bits to be extracted from each

pixel.

Step 3: Obtain the secret data sub-streams as by taking the

difference between above calculated difference values and

respective lower bounds. Transform to binary strings with

length equivalent to .

Finally, concatenate , to obtain the original bit sequence

of the secret message.

Lower-level Higher-level

=[0,7] =[8,15] =[16,31] =[32,63] =[64,255]

Fig.1 The dividing case (div=31) of the proposed method with lower level and higher level.

NaveenTypewritten textInt. Journal of Electrical & Electronics Engg. Vol. 2, Spl. Issue 1 (2015)e-ISSN: 1694-2310 | p-ISSN: 1694-2426

NaveenTypewritten textNITTTR, Chandigarh EDIT -2015 140

III. EXPERIMENTAL RESULTS

The simulation is done u