Data Hiding In A Binary Image

Ahmed Al-Jaber and Khair Eddin Sabri Department of Computer Science King Abdullah II School for Information Technology (KASIT) University of Jordan Email: [email protected] , [email protected]

Abstract In this paper, a new algorithm based on the CPT-algorithm [1] has been proposed for data hiding in a binary image. This technique does not need any key for embedding and extracting data. Also, it allows hiding four bits in a block of size 55 by changing a maximum of two bits. The proposed algorithm ensures security and the hiding effect is quite invisible. A comparison with the CPT algorithm is also described.

Keywords: steganography, watermarking, binary images, security, data hiding. 1. Introduction The security of digital media becomes of major concern due to its emergencies and wide spread. The security of the transformation of hidden data can be achieved by two ways: encryption and steganography. A combination of the two techniques can be used to further increase the security of data. In encryption, the message is changed so that no data can be disclosed if received by an attacker [2]. While in steganography, the secret message is embedded into an image (or any media) called cover image, and then sent to the receiver who extracts the secret message from the cover message [3]. After embedding the secret message, the cover image is called a stego-image. This image should not be distinguishable from the cover image, so that the attacker cant discover any embedded message. There are many techniques for encrypting data [2], which vary in their security, robustness, performance and so on. Also, there are many ways for embedding a message into another one. The most popular one is embedding a message into a colored image using LSB [4]. This method hides the data in the least significant bit of each pixel in the cover image. However, there are other known methods for hiding messages [4,5,6,7]. In the proposed scheme, the binary (black and white) image is used for hiding the secret message. Hiding in a binary image is very difficult due to the fact that

changing one bit in a binary image is easy to detect as it changes the color from black to white or the opposite. However, there are many techniques for a hiding a message in a binary image [1,8,9]. This paper presents a new scheme for embedding four bits in 55 block in a binary image by changing at most two bits with an efficient and invisible way without using a key. Finally, it should be noted that steganography is different from watermarking technique [10]. Watermarking is not used to transmit a secret message but to embed data, which might be visible, to guarantee ownership. The rest of the paper presents the CPT algorithm in the 2nd section, the proposed method in the 3rd section, experimental results in the 4th section, and the conclusion in the 5th section.

2. Review of an Existing Scheme The CPT-algorithm[1] is a well-known algorithm applied on a binary cover image to hide data. This algorithm is designed to embed at most log(mn+1) bits of the secret message in a mn block by changing a maximum of two bits. Before discussing the algorithm, the following notations are defined: K: Secret key used to embed and extract the message. It has a size of mn, and contains 0 & 1. W: Weight matrix also used in embedding and extracting the message. Its size is mn and its numbers range from 1 to 2r 1. r: Number of bits embedded in one block. F: Covered image used to hide data.

The Embedded Algorithm Input: Cover image (F), secret key (K), weight matrix (W) and secret message. Output: Stego-image. Step1: Divide F into blocks (Fi), each of size mn. Step2: Determine the size of r where r log(mn+1) Step3: Perform the following steps for each Fi until the whole secret message is embedded: a) Find (Fi K) W b) Find the sum (s) of the matrix obtained by the previous step. c) Find s mod m where m=2r d) Get the sequence of bits from the secret message. e) Compare the values obtained from c and d: if they are equal, no action is required, else, one or two bits in Fi should be changed to make the values equal.

The Extracted Algorithm Input: Stego-image, secret key (K) and weight matrix (W).

Output: Secret message. Step1: Divide F into blocks (Fi), each of size mn. Step2: Determine the size of r. Step3: Perform the following steps for each Fi until the whole secret message is extracted: a) Find (Fi K) W b) Find the sum (s) of the matrix obtained by the previous step. c) Find s mod m where m=2r. The result is r bits of the secret message. However, the CPT algorithm can embed 4 bits in a 55 block using a secret key and a weight matrix. The proposed algorithm takes a fixed block size 55 that can embed 4 bits by changing a maximum of 2 bits as in the CPT algorithm without using a secret key or a weight matrix as described in the following section. 3. Proposed Algorithm This algorithm is based on dividing the cover image into blocks, each of size 55. In each block, 4 bits from the secret message can be embedded with changing at most two bits in the block. Embedding Algorithm Input: Cover image and secret message Output: Stego-image. Step1: Divide the cover image into blocks (F) each of size 55. Step2: For each block, except single-value black and white ones, proceed as follows: 1. For every row in the first four rows of the block, exclusive-or all the bits of that row to get r1r2r3r4. 2. For every column in the first four columns of the block, exclusive-or all the bits of that column to get c1c2c3c4. 3. Exclusive-or the results in 1 and 2 to get s1s2s3s4 where s1=r1c1, s2=r2c2, and so on. 4. Compare the result obtained from 3 with the four embedded bits b1b2b3b4. If there is no difference, no change of bits in F is needed, otherwise, consider the following cases: if the difference in one bit bi, the bit [F] i,5 or [F]5,i should be changed

else if difference in two bits bi and bj, then the bit [F] i,j or [F]j,i should be changed. else if difference in three bits bi, bj and bk, then the bits (( [F] i,j or [F] j,i) and ( [F]k,5 or [F]5,k)) or (( [F] i,5 or [F] 5,i) and ( [F]k,i or [F]j,k)) or (( [F] 5,j or [F] j,5) and ( [F]k,i or [F]i,k)) should be changed

else (difference in four bits bi, bj, bk and bm) then the bits (( [F] i,j or [F] j,i) and ( [F]k,m or [F]m,k)) or (( [F] i,m or [F] m,i) and ( [F]k,i or [F]j,k)) or (( [F] m,j or [F] j,m) and ( [F]k,i or [F]i,k)) should be changed

The selection of the bit depends on the number of adjacent bits with the same value. The bit that has the least number of adjacent bits is selected. This is because it has a minimum effect on the cover image when it is changed. Extracting Algorithm Input: Stego-image Output: Secret message The algorithm used for extracting is similar to that used for embedding. It performs the following steps to give the embedded data. Step1: Divide the cover image into blocks (F) each of size 55. Step2: For each block, except single-value black and white ones, proceed as follows: 1. For every row in the first four rows of the block, exclusive-or all the bits of that row to get r1r2r3r4. 2. For every column in the first four columns of the block, exclusive-or all the bits of that column to get c1c2c3c4. 3. Exclusive-or the results in 1 and 2 to get the embedded bits s1s2s3s4 where s1=r1c1, s2=r2c2, and so on.

Example: Suppose that we have the following cover image: 1 1 1 1 1 1 0 0 1 0 1 1 0 0 0 1 0 1 0 1 0 1 0 1 0 0 0 0 0 0 1 0 1 0 0 1 1 0 1 0 1 1 1 1 1 1 1 0 0 0 and the embedded data: 11100101

Applying the step For every row in the first four rows of the block, exclusive-or all the bits of that row to get r1r2r3r4 on the given block, the following result is obtained: 11111=1 10100=0 11000=0 The result is 1000 01010=0

Applying the step For every column in the first four columns of the block, exclusive-or all the bits of that column to get c1c2c3c4 on the same block to get the following result: 11011=0 10101=1 11101=0 The result is 0011 10011=1

Applying the step Exclusive-or the results in 1 and 2 to get the embedded bits s1s2s3s4 where s1= r1c1, s2=r2c2, and so on on the same block to get the following result: 10=1 00=0 01=1 01=1

The result is 1011 and the embedded data is 1110.

As we can see, the bits number 2 and 4 in the result are different form those in the embedded data. So there should be a change in either the bit [F]2,4 or [F]4,2. Considering its effect on the cover image, the bit [F]4,2 should be changed. In the other block, if the same operation is repeated, the following results are obtained: r1r2r3r4 = 0101 c1c2c3c4 = 0010 s1s2s3s4 = 0111

The embedded data is 0101.

Therefore, the bit that should be changed is either [F]3,5 or [F]5,3. Based on its effect on the cover image, [F]5,3 is to be changed.

4. Experimental Results The proposed algorithm together with the CPT algorithm have been applied on different images for different data. The following computations were performed for each stego-image: Similarity: Between the stego-image and the original image Average: It is computed for each pixel depending on its neighbors. Then the average of pixel average values is also computed to test the consistency between each pixel and its neighbors. Standard Deviation: Compute the average for each pixel depending on its neighbors, and then compare it with the original image.

The result of these factors on Lena image is shown in figure 1, figure 2 and figure 3. Also, a stegoimage generated using the proposed algorithm and the CPT algorithm is shown in figure 4. A comparison of the proposed algorithm with the existing scheme shows that the proposed algorithm gives a better performance of the average and standard deviation factors, although CPT algorithm shows a better similarity factor. .
.
90.8

Lena 128x128
12

Lena 128x128
100 99.8

Lena 128x128

90.6

10

99.6

Similarity

Average

90.4 8

99.4 99.2 99 98.8 98.6 98.4

STD

90.2

90

4 89.8 2
0 50 100 150

89.6

98.2 0 50 100 150

Character
Proposed Algorithm CPT Algorithm

0 0 50 100 150

Character

character
Proposed Algorithm CPT Algorithm

Proposed Algorithm CPT Algorithm

Figure 1: Average comparison between the proposed Algorithm and CPT algorithm

Figure 2: "Standard Deviation comparison between proposed algorithm and CPT algorithm.

Figure 3: Similarity comparison between the proposed Algorithm and CPT algorithm.

(a)

(b)

(c)

Figure 4: (a) The original image, (b) The proposed algorithm, (c) CPT algorithm

5. Conclusions In this paper, a new technique has been proposed to hide data in a binary image. The used algorithm is secure and the hidden information is quite invisible. The advantages of the proposed technique are: The new algorithm shows a better performance of both the average and standard deviations between the original image and stego-image factors as we consider the neighbors of the modified bit. The new algorithm does not need a secret key, it needs only an agreement between the embedding and extracting agents. Whereas this key is necessary for the CPT algorithm, and the size of the key increases as the block size gets larger. In CPT algorithm, the weight matrix is needed to perform the multiplication of the three matrices to calculate the sum(S). This operation generates a large computational time. Whereas, in the proposed algorithm such computational time does not exist. A modification to this algorithm can be made to increase its security. For example, the block can be chosen randomly rather than sequentially depending on a seed number. Also, the ignored row and column of an order other than 5 can be changed so that the attacker cant get the message. References [1] Y. Chen, H. Pan, and Y. Tseng, A secure Data Hiding Scheme for Two-Color Images, IEEE Symposium On Computers and Communications, 2000 [2] A. Menezes, P. Oorschot, S. Vanstone, Handbook of Applied Cryptography CRC Press, 1996. [3] N. Johnson and S. Jajodia, Exploring steganography: seeing the unseen, IEEE Computer, pp. 26-34, February 1998.

[4]

K. Stefan and A. Fabien Information hiding techniques for steganography and digital watermarking, Artech House, 2000

[5]

F. Petitcolas, R. Anderson and M. Kuhn Information Hiding, A Survey Proceedings of the IEEE, special issue on protection of multimedia content, July 1999.

[6]

W. Bender, D. Gruhl, N. Morimoto, and A. Lu, Techniques for data hiding, IBM Systems Journal, Vol. 35, No. 3 and 4, pp. 313-336, 1996.

[7]

M. Celik, G. Sharma, and A. M. Tekalp, Reversible data hiding Proceeding IEEE ICIP, Rochester, NY, Sept. 2002.

[8]

M. Wu, E. Tang, B. Liu: "Data Hiding in Digital Binary Image", IEEE International Conference on Multimedia & Expo (ICME'00), New York City, 2000.

[9]

M. Wu and J. Lee, "A Novel Data Embedding Method for Two-Color Facsimile Images", Proeeding of International Symposium on Multimedia Information Processing, Taiwan, December 1998.

[10] C. Podilchuk and E. Delp, "Digital Watermarking: Algorithms and Applications," IEEE Signal Processing Magazine, Vol. 18, pp. 33-46, 2001.