Image Segmentation to Secure LSB2 Data Steganography

A digital color image usually has a high resolution, thus its size is good enough and the image can be used as a covering (holding) image to hide secrete messages (short and long). The methods commonly used for data steganography, e.g. LSB and LSB2 are not secure, so in this paper, a method of securing the LSB2 method is proposed. The proposed method is based on wavelet packet decomposition. The levels of decomposition will be kept in secret and one of the resulting segments will be used as a covering segment. MSE, PSNR, hiding time, and extraction time will be experimentally analyzed to prove that the proposed method is capable of handling the process of hiding secret messages, either sort or long. Keywords-steganography; LSB2; MSE; PSNR; hiding time; extraction time; WPT; decomposition level; segment; security

INTRODUCTION Data steganography [1][2][3] is the process of hiding secret data into covering data. The covering data must be large enough in order to be capable to hide the secret data [4][5]. Data steganography [6][7] must provide the following important features [8]: • The changes in the holding data must not affect them while the concealment process result must not be visible to the naked eye [2].
• The Mean Square Error (MSE) [9] between the original covering data and the holding data must very small and close to zero.
• The Peak Signal-to-Noise Ratio (PSNR) [10][11] between the original covering data and the holding data must very big in order to keep the quality of the holding data high.
• The secret data hiding time must be minimal.
• The secret data extraction time must be minimal.
• The hiding method must be secure and the process of hacking must be very complicated.
• The method must be simple to implement.
• The method must be capable of hiding secret data of various sizes (short and long messages).
One of the most common types of data that can be used to hide confidential messages is digital color images for the following reasons [12][13][14][15]: • The wide spread use of digital images.
• The sheer volume of covering data that a digital image provides [16,17].
• The possibility of reshaping the image before the process of masking data.
• The possibility of using a section of the image to implement the concealment process [20].

II. HIDING DATA METHODS
One of the most popular methods of data hiding is the Least Significant Bit (LSB) method which requires 8 bytes from the holding image to hide one character from the secret message. The LSB2 method is a modification of the LSB but it doubles the capacity of hiding by using 4 bytes from the covering image to hide one character from the secret message. The least two significant bits are used to hold data from the secret message as shown in Table I. The LSB2 adds minor changes to the covering image, ranging from +3 to -3. These changes in the pixel colors cannot be noticed by the human eye. The process of data hiding and data extracting using the LSB2 method is very simple, Figure 1 shows the process of hiding, while Figure 2 shows the process of data extracting. The LSB2 method adds minor changes to the covering image. These changes cannot be noticed by the human eyes, thus this method keeps the holding image very close to the covering one, and minimizes MSE and maximizes PSNR between the covering and the holding images. As we can see in   III. AIM OF THE STUDY LSB2 is a method of secret data is an easy-to-implement and quick-to-perform way, but one of its main flaws is its lack of security in the data-stripping device, due to its ease of penetration from non-authorized parties. Accordingly, the aim of this research is to update this method by strengthening it with the required protection operations and thus to prevent intruders from the possibility of obtaining or knowing the secret messages included in the digital image, provided that the advantages of the concealment method are preserved and without negatively affecting the efficiency of the method.

IV. RESEARCH METHOD
The hiding process is going to be implemented in four phases. The information in the first two phases must be kept confidential in order to secure the data.
• Color image rearrangement. The color channels are rearranged, then the color image 3D matrix is reshaped into a one row matrix. The reshaping can be done either rowwise or column-wise.
• Row matrix decomposition. The principles of wavelet packet tree decomposition [21,22] are used to decompose the image row matrix. In this phase, we have to select the number of levels needed to divide the image into segments, and then we have to select the segment [23] where we must hide the secret massage ( Figure 5).
• The LSB2 method of data hiding is applied.
• The holding image is rearranged back.
The extraction process will be implemented in 3 phases.
• Image rearrangement: Here we have to use the information used in the hiding process.
• After we get the number of decomposition levels and the segment number, image decomposition is applied.
• The LSB2 method to extract the message from the selected segment is applied. V. RESULTS AND DISCUSSION Twelve images were processed, and each of them was rearranged by replacing the color channels from red, green, and blue to blue, red, and green. Each image matrix was reshaped from 3D form to 1D column-wise. The number of the selected decomposition levels was defined as 7, and segment 6 was selected for message hiding. Figure 6 shows the segments of one image.

www.etasr.com
Zaini: Image Segmentation to Secure LSB2 Data Steganography   3  37  113  37713  1  3  38  1  19494  2  3  90  1  129600  3  3  267  803  1285200  4  3  245  245  1081553  5  3  41  41  30567  6  3  90  1  129600  7  3  45  137  37744  8  3  45  137  37744  9  3  50  50  37839  10  3  150  1  472500  11  3  286  1  1529814  12 The obtained segments for each image are of different sizes and locations and when the decomposition level changes the segments, their sizes, and their locations also changed. Tables  II and III show the image segment information after applying 7 image decomposition levels. Segment 6 was chosen in each color image and a message with a length of 50 characters was selected and hided in each covering image. Table IV shows the obtained experimental results. We can notice the following facts: • The quality of the holding images is very high, the MSE value is very low, while the values of PSNR are very high.
• The hiding and extraction times are minimal.
• Increasing the image size leads to increased PSNR values as shown in Figure 7.
• If the size of one segment does not meet the message length, we can use more segments.
• It is very difficult to know the segment number and the segment size without knowing the decomposition levels. • The quality of the holding image is close to the quality of the covering image.
• Increasing the message length will lead to decreasing PSNR as shown in Figure 10.
• Increasing message length will lead to increased hiding time as shown in Figure 11.
• Increasing message length will lead to rapidly increasing extraction time, as shown in Figure 12.    The obtained experimental results showed that the proposed method can be recommended to be used instead of the LSB2 method, because it can add a security level without affecting the efficiency and capacity of the LSB2 method. From the obtained results we can see that the quality parameters are better when we use images with big sizes.

VI. CONCLUSION
In this paper, a secure LSB2 method of data steganography was proposed, implemented, and analyzed. The obtained experimental results showed that selecting the image rearrangement method, the number of decomposition levels, and a segment number will increase the security level of the LSB2 method. Using image segments for data hiding will keep the parameters of the LSB2 method without negative changes. The values for MSE, PSNR, hiding time, and extraction time remain optimal.