CN109660693B - Robust steganography method for JPEG images based on the relative relationship between serial concatenated codes and DCT coefficients - Google Patents

Abstract

The present invention relates to Information Hiding Techniques fields, more particularly to the jpeg image robust steganography method based on serial concatenation of codes and DCT coefficient relativeness.Jpeg image robust steganography method based on serial concatenation of codes and DCT coefficient relativeness, comprising: determine modifier area；Design corresponding distortion function；Message to be embedded is encoded using Error Correction of Coding；Error Correction of Coding is embedded in using robust steganography；Carry out JPEG compression；It is obtained using the decoding of steganography containing wrong information flow, and carries out error correction using the decoding of Error Correction of Coding；Obtain the message of original insertion.The present invention can be in the wider middle JPEG compression attack for resisting the different quality factor, and can guarantee that insertion message can be extracted completely with high probability, meanwhile, in the case where carrying same secret message, the present invention has stronger anti-detection.

Description

Robust Steganography of JPEG Images Based on Relative Relationship between Serial Concatenated Codes and DCT Coefficients method

technical field

The invention relates to the technical field of information hiding, in particular to a robust steganography method for JPEG images based on the relative relationship between serial concatenated codes and DCT coefficients.

Background technique

Steganography is a technology that uses public carriers to construct covert communication channels. In recent years, with the development of mobile communication technology, more and more digital carriers are sent to and received from smart mobile terminals. However, due to limitations of bandwidth, tariffs, traffic, etc., terminals often damage the carriers transmitted on such channels. compression, so that classical steganography cannot correctly extract embedded messages in such cases. Therefore, it is of high practical value to study a steganographic algorithm that can completely extract and embed messages in such channels and has stronger anti-steganographic detection performance than robust watermarking algorithms.

Compared with the traditional JPEG adaptive steganography algorithm, the existing robust steganography algorithm DCRAS (DCT Coefficient Relationship Based Adaptive Steganography) has greatly improved the performance of resisting JPEG compression, and has stronger detection resistance than the robust watermarking algorithm. . However, when the algorithm resists JPEG compression, when the quality factor of JPEG compression attack is different from that of the carrier image, it is difficult to extract the secret message completely and correctly. In this case, the extracted message has about 1% error, which leads to For some practical application problems, how to reduce this error rate, ensure that the message can be extracted completely and correctly with a high probability, and at the same time improve its security, that is, anti-detection, is a problem that needs to be solved in robust steganography.

SUMMARY OF THE INVENTION

Aiming at the above problems to be solved by robust steganography, the present invention proposes a robust steganography method for JPEG images based on the relative relationship between serial concatenated codes and DCT coefficients, which improves the correct extraction rate and security of embedded messages.

In order to achieve the above object, the present invention adopts the following technical solutions:

A robust steganography method for JPEG images based on the relative relationship between serial concatenated codes and DCT coefficients, including the following steps:

Step 1: Determine the modification area;

Step 2: Design the corresponding distortion function;

Step 3: use error correction coding to encode the message to be embedded;

Step 4: Use robust steganography to embed error correction coding;

Step 5: JPEG compression;

Step 6: utilize steganographic decoding to obtain an error-containing information stream, and use the decoding of error correction coding to correct errors;

Step 7: Get the original embedded message.

Preferably, the step 1 includes:

Step 1.1: Decode the JPEG image to obtain blocks of DCT coefficients;

Step 1.2: correspond the blocks of the DCT coefficients to the elements of the virtual steganographic carrier;

Step 1.3: Establish a robust channel.

Preferably, the step 2 refers to designing corresponding distortion functions for the elements of the virtual steganographic carrier.

Preferably, the error correction code is a serial concatenated code, and the serial concatenated code is composed of an outer code and an inner code.

Preferably, during encoding, the serial concatenated code is performed in the order of encoding the outer code first and then the inner code; during decoding, it is performed in the reverse order of encoding.

Preferably, the outer code uses a convolutional code and the inner code uses RS coding when designing a serial concatenated code.

Preferably, the step 3 includes:

Step 3.1: use (3, 2, 5) convolutional code with a code rate of 2/3 to perform outer encoding on the message to be embedded to obtain an outer code word, and puncture the output outer code word;

Step 3.2: perform an interleaving operation on the punctured outer code word;

Step 3.3: Use (200, 150, 8) RS coding with a code rate of 3/4 to perform inner coding on the outer code word after the interleaving operation.

Preferably, the decoding of the convolutional code is a Viterbi decoding algorithm, and the code depth is 25.

Preferably, the coding of the RS code is minimum distance coding, and the code depth is 25.

Compared with the prior art, the present invention has the following beneficial effects:

The present invention proposes a new error correction code for robust steganography. The code consists of two parts: a convolutional code and an RS code, wherein the convolutional code is an outer code, the RS code is an inner code, and an interleaver is used between the inner and outer codes. By cascading and setting appropriate encoding parameters, the improved robust steganography algorithm can resist JPEG compression attacks with different quality factors in a wider range.

Aiming at the problem of how to further improve the compression resistance and detection resistance of the robust steganography algorithm, the invention proposes a robust steganography method for JPEG images based on the relative relationship between serial concatenated codes and DCT coefficients on the basis of the robust steganography algorithm , by designing serial concatenated codes, using the characteristics of convolutional codes and RS coding, and combining with puncturing technology, a serial concatenated code with high error correction capability and high code rate suitable for robust steganography is proposed. And the robust steganography algorithm is improved on its basis. The present invention can ensure that the embedded message can be completely extracted with high probability. At the same time, under the condition of carrying the same secret message, the method proposed by the present invention has stronger performance. detection resistance.

Description of drawings

FIG. 1 is one of the basic flow charts of the robust steganography method for JPEG images based on the relative relationship between serial concatenated codes and DCT coefficients according to the present invention.

FIG. 2 is the second schematic flow chart of the basic flow chart of the robust steganography method for JPEG images based on the relative relationship between serial concatenated codes and DCT coefficients according to the present invention.

FIG. 3 is a schematic diagram of the structure of the serial concatenated code of the JPEG image robust steganography method based on the relative relationship between the serial concatenated code and the DCT coefficients of the present invention.

FIG. 4 is a schematic structural diagram of a convolutional code of a robust steganography method for JPEG images based on the relative relationship between serial concatenated codes and DCT coefficients according to the present invention.

Detailed ways

For ease of understanding, the following explanations are made to some terms that appear in the specific embodiments of the present invention:

RS coding: It is a kind of linear grouping coding based on the finite field unary polynomial proposed by Irving S. Reed and Gustave Solomon.

The present invention will be further explained below in conjunction with the accompanying drawings and specific embodiments:

Example 1:

As shown in Figure 1, a JPEG image robust steganography method based on the relative relationship between serial concatenated codes and DCT coefficients of the present invention comprises the following steps:

Step S101: determine the modification area;

Step S102: Design a corresponding distortion function;

Step S103: use error correction coding to encode the message to be embedded;

Step S104: using steganography to embed error correction coding;

Step S105: perform JPEG compression;

Step S106: use the decoding of error correction coding to obtain an error-containing information stream, and perform error correction;

Step S107: Obtain the original embedded message.

Embodiment 2:

As shown in Figure 2, another JPEG image robust steganography method based on the relative relationship between serial concatenated codes and DCT coefficients of the present invention comprises the following steps:

Step S201: Determine the modification area, including:

Step S2011: Decode the JPEG image I to obtain a series of 8×8 blocks D _k ={D _k (i),i=1,2,...,64} of the quantized and rounded DCT coefficients D, where , k=1,2,...,B, B is the total number of blocks of 8×8 image blocks, D _k (i) represents the ith coefficient in the kth quantized DCT block;

Step S2012: Use the DCT coefficient D _k (i) to use the DCT coefficient of the same position in the adjacent block as the adjacent block mean The relative relationship of the 4 adjacent DCT coefficients Corresponds to a virtual steganographic carrier of length l ₁ On the element x _j on , 1≤j≤l ₁ , satisfies:

Step S2013: Establish a robust channel:

Establish a robust channel on the basis of X as a carrier, that is, determine how to embed X into the message and get the carrier secret Map to the original vector {D _k } _1≤k≤B :

To determine the value of σ _ki in formula (2), set And T ₁ , T ₂ are given thresholds and T ₁ <T ₂ , then the parameter σ _ki is determined by the following formula:

Among them, T ₁ =15, T ₂ =30, get the parameter set when 0 is embedded and the parameter set when embedding 1

Step S202: Design a corresponding distortion function for the elements of the virtual steganographic carrier:

The distortion function of the J-UNIWARD algorithm with strong anti-detection performance Combined with the proposed robust channel, and knowing the quality factor qf of the JPEG compression attack suffered in advance, the carrier image IMG can be compressed once using JPEG compression with a quality factor of qf, and after compression, we get Its image DCT coefficients 8x8 blocks of and the neighborhood mean Using this preprocessing and Design the distortion function DFR(x _j ,y _j ) on X:

in, and J-UNIWARD distortion values of magnitude 1 and 2 modified by the DCT coefficients D _k (i) after embedding the message, respectively And wet_cost represents the wet point distortion value, which is usually set to a large value, so that the STC code does not modify this point as much as possible during embedding. Then, the distortion function defined by equation (4) can be used to embed the secret message on the robust carrier X with STC encoding to obtain the secret object Y, and the carrier can be actually modified by equation (2), so as to obtain the embedded secret message. A dense carrier that is resistant to JPEG compression.

Step S203: Use error correction coding to encode the message to be embedded, the error correction coding is a serial concatenated code, and the serial concatenated code is composed of two parts: an outer code and an inner code; the basic structure of the serial concatenated code As shown in Figure 3, wherein, during encoding, the serial concatenated code is performed in the order of first encoding the outer code and then encoding the inner code; during decoding, it is performed in the reverse order of encoding and encoding; Step S203 further includes:

Step S2031: Use (3, 2, 5) convolutional code with a code rate of 2/3 to perform outer encoding on the message to be embedded to obtain an outer code word, and puncture the output outer code word, as follows:

A convolutional code can also be expressed as (n, k, t), where n is the bit length of the data packet output by the encoder, k is the bit length of the data packet input by the encoder, and t is the convolutional code constraint length.

The secret message bit sequence to be embedded msg _i ∈{0,1} is divided into groups of 2 bits. If l ₂ is not an even number, add a 0 bit at the end to obtain the grouped message sequence where cm _i =(msg _2i-1 , msg _2i ).

The outer code used in the construction is a (3,2,5) convolutional code with a typical code rate of 2/3, and the packet message sequence cm is encoded to obtain the outer code codeword sequence l ₄ =3l ₃ , and the structure of the convolutional code is shown in FIG. 4 .

It can be seen that the characteristic of the convolutional code is that the output codeword is not only related to the current input information bit group, but also to the previous t-1 input information bit group, and its encoding characteristic is iterative. At the same time, the convolutional code encoder will output a 3-bit convolutional code word for every 2 bits of information input, which means that the code rate of the serial concatenated code is 2/3. In order to further improve the code rate, use the puncturing matrix PM to puncture the encoded codeword:

The three columns in the matrix PM respectively represent the three groups of code words (oc _3i-2 , oc _3i-1 , oc _3i ) ^T , i=v, v+1, v+2 in the outer code code word sequence oc how the output bits are To choose, group the output bit sequence of the encoder according to the number of columns of PM before punching. If the element in the matrix is "1", it means that the corresponding output bit is reserved, and if the element in the matrix is "0", it means that the corresponding output bit is reserved. Output bits are truncated. The design idea of matrix PM is: oc _3i-2 is the original sequence element, so all the sequence elements should be reserved when punching; at the same time, oc _3i-1 and oc _3i are used as the output elements of convolution iteration information, so when punching At the same time, the elements in oc2 and oc1 cannot be discarded at the same time, that is, the same column of the next two rows of the punching matrix is not "0" at the same time. When the code sequence obtained by the outer encoder is punctured according to formula (5), the punctured outer code codeword sequence oc _{punc is obtained} , and the code rate of the outer code becomes (6/9)×(9/7) =6/7.

Step S2032: After obtaining the outer code word oc _punc , perform an interleaving operation to reduce the correlation between the inner and outer codes and improve the performance of the concatenated code. The interleaver used is random interleaving, that is, controlled by a key seed. The pseudo-random generator of scrambles the outer coding sequence randomly to obtain the scrambled outer coding codeword ioc _punc , and then performs inner coding.

Step S2033: Use (200, 150, 8) RS coding with a code rate of 3/4 to perform inner coding on the outer code word after the interleaving operation, as follows:

Denote the coding parameters of RS coding as (N, K, M), where N is the bit length of the data packet output by the encoder, N≤2 ^M , K is the bit length of the data packet input by the encoder, and M is GF(2 ^M ) of the order of the primitive polynomial, the code rate is K/N, set the parameter The parameters used are (200, 150, 8), Galois Field GF(2 ^M ), RS coding with a code rate of 3/4 on M=8, the primitive polynomial PRM(x of Galois Field GF(2 ^M ) ) is set to:

PRM(x)= ^x8 + ^x4 + ^x3 + ^x2 +1(6)

RS coding has 2 ^M -1=255 code words, each of which can be regarded as an element in Galois finite field GF(2 ^M ), and the construction of the generator polynomial g(x) of RS coding for:

g(x)=(x+α ¹ )(x+α ² )…(x+α ^2T )=x ⁵⁰ +g ₄₉ x ⁴⁹ +…+g ₁ x+g ₀ (7)

where α is a prime element in GF(2 ^M ), _{gi ∈} GF(2 ^M ), i=0,1,...,49.

After completing the above-mentioned RS encoding pre-operation, perform actual RS encoding on the outer code word ioc _punc after the interleaving operation: firstly, the ioc _punc is grouped according to the parameter K=150 of RS encoding (zero-filling at the end of the code word enables it to be Divisible by K) to obtain the packet sequence to be encoded by RS The corresponding information polynomial is:

The encoded code word is composed of K=150 information symbol symbols in the front, and NK=50 correction symbols in the back, and in order to obtain its corresponding polynomial, first the information code, that is, the polynomial m ^{( i)} (x) upgrade NK=50 order, namely multiply by x ⁵⁰ ; then divide with the generator polynomial g(x) to obtain the remainder r(x); finally calculate x ⁵⁰ m ⁽ⁱ⁾ (x)+r(x ) is the polynomial corresponding to the encoded codeword, and the longest continuous error bit length that it can resist is T=25.

Step S204: using robust steganography to embed error correction coding;

As an implementation manner, STC coding is used to steganographically embed error correction coding.

Step S205: Open the channel for JPEG compression.

Step S206: use STC decoding to obtain an error-containing information stream, and use the decoding method of error-correcting codes to correct errors:

The decoding of the outer code, that is, the decoding of the convolutional code, is the Viterbi decoding algorithm, and the code depth is 25; the decoding of the inner code, that is, the decoding of the RS code, is the minimum distance decoding, and the code depth is 25;

The decoding is performed according to the order of decoding the inner code first and then decoding the outer code, that is, the order of decoding the inner code first and then decoding the outer code.

Step S207: obtain the original embedded message: obtain the embedded message according to the STC code extraction algorithm;

The image after JPEG compression is recorded as The image DCT coefficient, block and neighborhood mean are respectively recorded as and The extracted carrier is recorded as Its extraction rules are as follows:

The above are only the preferred embodiments of the present invention. It should be pointed out that for those skilled in the art, without departing from the principles of the present invention, several improvements and modifications can be made. It should be regarded as the protection scope of the present invention.

Claims (6)

1. based on the JPEG image robust steganography method of serial concatenated code and DCT coefficient relative relation, it is characterized in that, comprise the following steps: Step 1: Determine the modification area; The step 1 includes: Step 1.1: Decode the JPEG image to obtain blocks of DCT coefficients; Step 1.2: correspond the blocks of the DCT coefficients to the elements of the virtual steganographic carrier; Step 1.3: Establish a robust channel; Step 2: Design a corresponding distortion function for the elements of the virtual steganographic carrier; Step 3: use error correction coding to encode the message to be embedded; Step 4: using robust steganography to embed error correction codes; the error correction codes are serial concatenated codes, and the serial concatenated codes are composed of two parts, an outer code and an inner code; Step 5: JPEG compression; Step 6: utilize steganographic decoding to obtain an error-containing information stream, and use the decoding of error correction coding to correct errors; Step 7: Get the original embedded message. 2. the JPEG image robust steganography method based on the relative relation of serial concatenated codes and DCT coefficients according to claim 1, it is characterized in that, during encoding, serial concatenated codes are encoded according to the encoding of the outer code first and then the inner code. Order; when decoding, follow the reverse order of encoding. 3. the JPEG image robust steganography method based on the relative relationship between serial concatenated codes and DCT coefficients according to claim 1, is characterized in that, when designing serial concatenated codes, the outer code uses convolutional codes and The inner code uses RS coding. 4. the JPEG image robust steganography method based on the relative relation of serial concatenated codes and DCT coefficients according to claim 1, is characterized in that, described step 3 comprises: Step 3.1: use the (3, 2, 5) convolutional code with a code rate of 2/3 to perform outer encoding on the message to be embedded to obtain an outer code word, and puncture the output outer code word; Step 3.2: perform an interleaving operation on the punctured outer code word; Step 3.3: Use (200, 150, 8) RS coding with a code rate of 3/4 to perform inner coding on the outer code word after the interleaving operation. 5. the JPEG image robust steganography method based on the relative relation of serial concatenated codes and DCT coefficients according to claim 3 or 4, is characterized in that, the decoding of convolutional code is Viterbi decoding algorithm, and the code The depth is 25. 6. the JPEG image robust steganography method based on the relative relation of serial concatenated codes and DCT coefficients according to claim 3 or 4, is characterized in that, the decoding of RS code is minimum distance decoding, and the code depth is 25.