Electronic document signing method based on fixed neural network steganography technology

Abstract

The invention discloses an electronic document signing method based on a fixed neural network steganography technology. The method comprises the following steps: a signer verifies the identity of a signer by using a trusted identity authentication platform, applies for a signature key by using an identity ID passing the verification, and performs abstract operation on a to-be-signed document through a hash function to generate a document plaintext abstract; generating a signature value by using the signature key and the document plaintext abstract through an identification password signature algorithm; taking the signature value, the document plaintext abstract and the verified identity ID as verification values, uploading the verification values and the signature picture of the signer to a cloud platform, and embedding the signature verification value of the identification password number into the handwritten electronic signature picture by the cloud platform by utilizing a fixed neural network steganography technology; according to the method, the digital signature information can be safely hidden in the picture on the premise that the visual appearance of the signature picture is not obviously changed, and the safety and the credibility in the electronic document processing process are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of digital signatures, and more specifically to an electronic document signing method based on fixed neural network steganography technology. Background Art

[0002] Handwritten signatures are mainly used to confirm the signer's approval and recognition of the content of the document, demonstrate the signer's willingness to agree, and in many cases have legal effect.

[0003] The existing handwritten electronic signature signing method mainly relies on PKI technology to ensure the security and reliability of handwritten electronic signatures, but there are still the following shortcomings:

[0004] Insufficient signatory identity verification: Existing handwritten electronic signature signing methods mostly verify the signing device or signing account through CA certificates, but lack verification of the user who actually operates the signing device or account. This increases the risk of attackers obtaining the signing account and device through impersonation or other illegal means.

[0005] Restrictions on the additional methods of digital signatures: Existing methods of signing handwritten electronic signatures mostly use digital signature technology to ensure the connection between handwritten signatures and documents. When verifying a digital signature, the digital signature value needs to be attached to the handwritten electronic signature and sent to the verifier. Existing additional methods are mainly divided into two categories: one is to restrict the signature scenario, such as restricting the format of the verified file and the verification environment. As a result, verification can only be completed in a specific environment, and once the file with the electronic signature is transferred to a copy of another file form, the verification function may be lost. The second is to attach the digital signature value to the document or handwritten electronic signature in a visible or invisible way through a special encoding method. The former mostly uses intuitive methods such as visible watermarks and QR code technology. These methods often require a certain amount of visible space on the document, which may interfere with the original layout and aesthetics of the document. This intuitive method is not suitable for some formal business or legal documents. The latter mostly uses image steganography technology, such as modifying the image pixel value or using deep learning technology to hide the digital signature value in the document or handwritten electronic signature. These methods cannot accurately restore the digital signature value, and the use of deep learning methods is also accompanied by huge overhead.

[0006] Certificate management issues: The certificate management process is complex and improper management may lead to security risks. Summary of the invention

[0007] In view of this, the present invention provides an electronic document signing method based on fixed neural network steganography technology, which is used to at least solve some of the technical problems in the background technology.

[0008] In order to achieve the above object, the present invention adopts the following technical solution:

[0009] The present invention discloses an electronic document signing method based on fixed neural network steganography technology, comprising the following steps:

[0010] Signatory identity registration:

[0011] The signer uses a trusted identity authentication platform to verify the signer's identity and uses the verified identity ID to apply for the signing key S k ;

[0012] The signatory uploads the signer's signature image C to the cloud platform;

[0013] Document signing:

[0014] The signer performs a digest operation on the document Doc to be signed using a hash function to generate a document plaintext digest H;

[0015] Using the signature key S k and document plaintext digest H, and generates a signature value M through the identification cryptographic signature algorithm;

[0016] The signature value M, the document plaintext summary H, and the verified identity ID are used as the verification value N and uploaded to the cloud platform;

[0017] The cloud platform uses fixed neural network steganography technology to embed the verification value N into the signature image C to obtain a steganographic image C′ containing the verification information N, and inserts the steganographic image C′ as a signature into the document Doc to be signed to obtain the signed document Doc′.

[0018] Furthermore, the above method also includes verifying the obtained signed document Doc′ to verify the security and validity of the signed document.

[0019] Furthermore, in a specific implementation, the obtained signed document Doc′ is verified and executed in the cloud platform, specifically including the following steps:

[0020] Extract the hidden image C′ from the signed document Doc′, and use the signed document Doc′ after removing the hidden image C′ as the verification document Doc″;

[0021] The decoder in the fixed neural network steganography FNNS model is used to extract the verification value N of the extracted stego-image C′;

[0022] Calculate the plaintext summary of the verification document Doc″ using the same hash function as the document signing step to obtain the plaintext summary value H′ corresponding to the verification document Doc″;

[0023] Compare the document plaintext digest H in the verification value N with the digest value H′ corresponding to the verification document Doc″ to see if they are consistent. If they are consistent, the signature document Doc′ has not been tampered with. If they are inconsistent, the signature document Doc′ has been tampered with.

[0024] When the signed document Doc′ has not been tampered with, a verification signature value M′ is generated by identifying a cryptographic signature algorithm based on the signatory identity ID extracted from the verification value N and the document digest value H in the verification value N;

[0025] Compare the signature value M in the verification value N with the verification signature value M′. If they are consistent, the document signature is valid; if they are inconsistent, the document signature is invalid.

[0026] Furthermore, in another embodiment, the obtained signed document Doc′ is verified, specifically comprising the following steps:

[0027] The cloud platform publishes the signed document Doc′ to a public website;

[0028] The client downloads the signed document Doc′ from the public website and transmits it to the cloud platform, and the client includes a signer and a verifier;

[0029] The cloud platform extracts the stego-image C′ from the signed document Doc′, and uses the signed document Doc′ after removing the hidden image C′ as the verification document Doc″;

[0030] The cloud platform uses the decoder in the fixed neural network steganography technology FNNS model to extract the verification value N of the extracted stego image C′, and returns the verification value N and the verification document Doc″ to the client;

[0031] The client calculates the plaintext digest of the verification document Doc″ using the same hash function as the document signing step, and obtains the plaintext digest value H′ corresponding to the verification document Doc″;

[0032] The client compares the document plaintext digest H in the verification value N with the digest value H′ corresponding to the verification document Doc″ to see if they are consistent. If they are consistent, the signature document Doc′ has not been tampered with. If they are inconsistent, the signature document Doc′ has been tampered with.

[0033] When the signed document Doc′ has not been tampered with, the client generates a verification signature value M′ based on the signatory identity ID extracted from the verification value N and the document digest value H in the verification value N by using the identification cryptographic signature algorithm;

[0034] The client compares the signature value M in the verification value N with the verification signature value M′. If they are consistent, the document signature is valid; if they are inconsistent, the document signature is invalid.

[0035] Furthermore, in the document signing step, the signer performs a digest operation on the document Doc to be signed through a hash function to generate a document plaintext digest H, which specifically includes the following steps:

[0036] The signer performs a digest operation on the document Doc to be signed using the SM3 algorithm to generate a document plaintext digest H.

[0037] Furthermore, in the document signing step, the cloud platform uses a fixed neural network steganography technology to write the verification value N into the signature image C to obtain a steganographic image C′ containing the verification value, which specifically includes the following steps:

[0038] The encoder in the fixed neural network steganography technology FNNS model is used to write the verification information N consisting of the signature value M, the document plaintext summary H, and the verified identity ID into the signature image.

[0039] Furthermore, the encoder in the fixed neural network steganography technology FNNS model is used to write the signature value M, the document plaintext summary H, and the verification information N composed of the verified identity ID into the signature image, which is achieved by optimizing the following minimum binary cross entropy loss function:

[0040]

[0041] Among them, L BCE represents the binary cross entropy loss function, C' is the stego-image containing verification information, N is the binary information consisting of the signature value M to be embedded, the document plaintext summary H, and the verified identity ID, and F is the preset fixed decoder network SteganoGAN.

[0042] Furthermore, the constraint condition for minimizing the binary cross entropy loss function is:

[0043] CC' ∞ ≤δ

[0045] The image change threshold caused by the δ embedding process can be known from the above technical solution. Compared with the prior art, the present invention discloses an electronic document signing method based on fixed neural network steganography technology, which has the following beneficial effects:

[0046] 1. The present invention combines a trusted identity authentication platform with an identification password digital signature algorithm to achieve identity authentication of the signer himself, thereby ensuring the authenticity of the document signing.

[0047] 2. The present invention uses the fixed neural network steganography technology FNNS to embed the digital signature value into the handwritten electronic signature of the signer. This not only ensures that the original layout of the document is not destroyed, but also ensures the accurate extraction of the digital signature value. FNNS reduces the computing cost and response time by fixing the trained neural network.

[0048] 3. The present invention can be easily extended to enterprises and institutions of all sizes and is applicable to various formal electronic documents that require signing, such as legal, financial and government documents. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0050] Figure 1 A schematic diagram of a document signing scenario and process provided by an embodiment of the present invention.

[0051] Figure 2 A flow chart of the signatory identity registration provided by an embodiment of the present invention.

[0052] Figure 3 A flowchart of document signing provided by an embodiment of the present invention.

[0053] Figure 4 A flow chart of signed document verification provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] The present invention discloses an electronic document signing method based on fixed neural network steganography technology. The goal of the present invention is to achieve the unique correspondence between the signer and the signed document, which mainly includes the steps of signatory identity registration and document signing. First, the identity of the signer is authenticated. After the identity authentication is passed, a signature value is generated by combining the signer's identity information and the signed document content by adopting an identification password signature algorithm, such as SM3 or SM9 identification password algorithm. Next, the signature value is embedded into the signer's handwritten signature image together with the verified identity ID and the document plaintext summary using fixed neural network steganography (FNNS), and the embedded handwritten signature image is input and output into the signed document to complete the document signing. Figure 1 The specific scenarios and processes of electronic document signing based on fixed neural network steganography technology are given.

[0056] As another implementation method, after completing the document signing, the signed document is also verified. During the verification process, the signature value is extracted from the signature image and a consistency check is performed to ensure the authenticity and consistency of the signature.

[0057] The applicable scenarios of the present invention include, but are not limited to, the signing of official documents, the signing of legal documents, the authentication of bank transactions, and any handwritten electronic signature application that requires security.

[0058] The method steps of the present invention are further described in detail below through more detailed embodiments.

[0059] The basic electronic document signing method based on fixed neural network steganography technology includes two parts: signatory identity registration and document signing, and on this basis also includes a document signing verification step.

[0060] The signatory identity registration is mainly completed on the cloud platform, such as Figure 2 As shown in the figure, the signer identity registration includes the following steps: First, the signer submits his or her identity information, including ID number, name, face information, phone number, email address, etc., to the trusted identity authentication platform through the signing party (signature device). After the trusted identity authentication platform authenticates correctly, it will feedback to the cloud platform. The cloud platform generates an identity ID (identity identifier) ​​that has passed the identity information verification, and transmits the authenticated identity ID to the key generation center KGC. The key generation center KGC uses the identity ID combined with the system signature master private key to generate the signer's signature key S k And return it to the signer. The signer submits the signer's handwritten signature image C to the cloud platform, which stores this data.

[0061] In the embodiment of the present invention, the identity-based signature algorithm (IBS) used in the signer identity registration step is a digital signature method based on user identity information. This technology omits the public key certificate in the traditional public key infrastructure (PKI), simplifies the key management process, and thus improves the convenience and efficiency of the operation. First, the key generation center (KGC) is responsible for setting the parameters of the entire system and generating a signature master key pair. The signature master private key is confidential and only held by KGC, and KGC discloses the signature master public key. Then the user needs to provide his identity ID to KGC. Based on the identity ID and the signature master private key of KGC, KGC generates the user's private key (signature key) and securely delivers it to the user. The user uses his private key to sign the message to be sent. This process usually involves processing the message content through a hash function, such as the SM3 encryption algorithm, and then using the private key to encrypt and generate a signature. When the message recipient receives the message with the signature, the sender's identity is used for verification.

[0062] Document signing is used to generate a signature. In the document signing step, the signer uses a hash function algorithm (such as SM3) to calculate the plaintext summary of the document to be signed Doc. When calculating the plaintext summary of the document to be signed, the content of the document to be signed Doc is first extracted, and then the plaintext summary of the extracted content of the document to be signed is calculated using the hash function algorithm. This process can be expressed by the following formula: H = Hash (Doc). In other embodiments, the document to be signed can also be in other forms such as PDF.

[0063] The signer uses the plaintext summary H of the document to be signed and the signature key S k And generate the signature value M=Sign(S k ,H), and upload the signature value M, document plaintext summary H, and verified identity ID as verification value N to the cloud platform. This process can be expressed as N = ID | H | M. The cloud platform embeds the verification data N into the signature image C through the encoder in the fixed neural network steganography technology FNNS model to obtain the steganographic image C′ = FNNS containing the verification value N. Encoder (C, N). The cloud platform inserts the steganographic image C′ as a signature into the document Doc to obtain the signed document Doc′.

[0064] The fixed neural network steganography FNNS model uses a preset neural network architecture to perform steganographic operations on images and embeds verification values ​​by optimizing a specific loss function. In this process, the original signature image C is first normalized to scale the pixel values ​​of the image to between 0 and 1 to match the input range of the neural network processing. The specific embedding process is completed by the neural network by solving the following minimum binary cross entropy loss function:

[0065]

[0066] Where C' is the image after the verification value is embedded, N is the binary verification value to be embedded, and F is a preset fixed decoder network. In this embodiment, the preset fixed decoder network uses SteganoGAN. The constraint condition of the loss function is CC' ∞ ≤δ, ensuring that the image change caused by embedding does not exceed the threshold δ to maintain the visual invariance of the image. 0≤C'≤1 ensures that the modified image pixel value is still within the valid pixel value range.

[0067] The entire embedding process is completed by optimizing the binary cross entropy loss function, which can measure the effect of embedding the verification value and guide the neural network how to accurately adjust the pixels. FNNS adds tiny perturbations to the image to embed information. These changes are almost undetectable visually and have little effect on the statistical properties of the image, making the steganographic content more hidden from steganalysis tools. FNNS can provide high concealment while ensuring high recovery rate and information integrity, and is suitable for steganographic communication scenarios with high security requirements.

[0068] The specific process of document signing can be Figure 3 The flowchart shown is described.

[0069] After completing the document signing, the signed document is also verified. A signature verification method is performed in the cloud platform and specifically includes the following steps:

[0070] Extract the hidden image C′ from the signed document Doc′, and use the signed document Doc′ after removing the hidden image C′ as the verification document Doc″;

[0071] The decoder in the fixed neural network steganography FNNS model is used to extract the verification value N of the extracted stego-image C′;

[0072] Calculate the plaintext summary of the verification document Doc″ using the same hash function as the document signing step to obtain the plaintext summary value H′ corresponding to the verification document Doc″;

[0073] Compare the document plaintext digest H in the verification value N with the digest value H′ corresponding to the verification document Doc″ to see if they are consistent. If they are consistent, the signature document Doc′ has not been tampered with. If they are inconsistent, the signature document Doc′ has been tampered with.

[0074] When the signed document Doc′ has not been tampered with, a verification signature value M′ is generated by identifying a cryptographic signature algorithm based on the signatory identity ID extracted from the verification value N and the document digest value H in the verification value N;

[0075] Compare the signature value M in the verification value N with the verification signature value M′. If they are consistent, the document signature is valid; if they are inconsistent, the document signature is invalid.

[0076] In addition to the above verification methods, there is another document signing verification method, the specific steps are as follows:

[0077] The cloud platform publishes the signed document Doc′ to a public website.

[0078] The client downloads the signed document Doc′ from the public website and transmits it to the cloud platform. The client here includes the signer (signature device) and the verifier (verification device).

[0079] The cloud platform extracts the steganographic image C′ from the signed document Doc′, and uses the signed document Doc′ after removing the hidden image C′ as the verification document Doc″, where the extracted steganographic image C′ contains verification information (verification value N).

[0080] The cloud platform uses the decoder Decoder in the fixed neural network steganography technology FNNS model to extract the verification value N of the extracted stego image C′, and returns the verification value N and the verification document Doc″ to the client. The verification value extraction process can be expressed as N=FNNS Decoder (C′).

[0081] The client calculates the plaintext digest of the verification document Doc″ (the signed document after removing the hidden image C′) using the same hash function as the document signing step, and obtains the plaintext digest value H′ corresponding to the verification document Doc″.

[0082] The client compares the document plaintext digest H in the verification value N with the digest value H′ corresponding to the verification document Doc″ to see if they are consistent. If they are consistent, the signed document Doc′ has not been tampered with. If they are inconsistent, the signed document Doc′ has been tampered with. In this process, the digest value H′ of the verification document can be expressed by the following formula H′=Hash(Doc′).

[0083] When the signed document Doc′ has not been tampered with, the client generates a verification signature value M′ through the identification cryptographic signature algorithm according to the signatory identity ID extracted from the verification value N and the document digest value H in the verification value N. The generation process of the verification signature value M′ is M′=Sign(ID,H);

[0084] The client compares the signature value M in the verification value N with the verification signature value M′. If they are consistent, the document signing is valid. If they are inconsistent, the document signing is invalid (proving that the signature image has been tampered with or the signature of the document has not been authenticated).

[0085] Among them, the verification steps of the signed document after the verification party (verification device) obtains the signed document Doc′ are as follows: Figure 4 As shown, the following steps are included:

[0086] The verifier sends the signed document Doc′ with the stego image C′ to be verified to the cloud platform. The cloud platform extracts the stego image C′ containing the verification information and submits it to the decoder in the fixed neural network steganography technology FNNS model.

[0087] The decoder in the FNNS model extracts the verification value N from the stego-image C′ and returns the verification value N and the signature document to be verified after removing the stego-image to the verifier.

[0088] The verifier uses the same hash function algorithm as in the document signing step to make a document summary for the signature document to be verified after removing the stego-image, and obtains the plaintext summary H′;

[0089] The verifier compares the plaintext digest H′ with the plaintext digest value H in the verification value N. If they are the same, the document has not been tampered with. If they are different, the document has been tampered with and the verification fails.

[0090] When the signed document Doc′ has not been tampered with, the verification signature value M′ is generated through the identification cryptographic signature algorithm based on the signatory identity ID extracted from the verification value N and the document digest value H in the verification value N; the verifier compares the signature value M in the verification value N in the steganographic image C′ with the calculation result M′. If they are the same, it means that the document signature is valid; if they are different, it means that the document signature is invalid.

[0091] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0092] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for signing electronic documents based on fixed neural network steganography technology, characterized in that: The following steps are included: Signatory identity registration: The signer uses a trusted identity authentication platform to verify the signer's identity and uses the verified identity ID to apply for the signing key S k ; The signatory uploads the signer's signature image C to the cloud platform; Document signing: The signer performs a digest operation on the document Doc to be signed using a hash function to generate a document plaintext digest H; Using the signature key S k and document plaintext digest H, and generate a signature value M through the identity cryptographic signature algorithm; The signature value M, the document plaintext summary H, and the verified identity ID are used as the verification value N and uploaded to the cloud platform; The cloud platform uses fixed neural network steganography technology to embed the verification value N into the signature image C to obtain a steganographic image C′ containing the verification information N, and inserts the steganographic image C′ as a signature into the document Doc to be signed to obtain the signed document Doc′.

2. According to claim 1, a method for signing electronic documents based on fixed neural network steganography technology is characterized in that: The method also includes verifying the obtained signed document Doc′ to verify the security and validity of the signing of the signed document.

3. According to claim 2, a method for signing electronic documents based on fixed neural network steganography technology is characterized in that: Verify the obtained signed document Doc′ and execute it in the cloud platform, which specifically includes the following steps: Extract the hidden image C′ from the signed document Doc′, and use the signed document Doc′ after removing the hidden image C′ as the verification document Doc″; The decoder in the fixed neural network steganography FNNS model is used to extract the verification value N of the extracted stego-image C′; Calculate the plaintext summary of the verification document Doc″ using the same hash function as the document signing step to obtain the plaintext summary value H′ corresponding to the verification document Doc″; Compare the document plaintext digest H in the verification value N with the digest value H′ corresponding to the verification document Doc″ to see if they are consistent. If they are consistent, the signature document Doc′ has not been tampered with. If they are inconsistent, the signature document Doc′ has been tampered with. When the signed document Doc′ has not been tampered with, a verification signature value M′ is generated by identifying a cryptographic signature algorithm based on the signatory identity ID extracted from the verification value N and the document digest value H in the verification value N; Compare the signature value M in the verification value N with the verification signature value M′. If they are consistent, the document signature is valid; if they are inconsistent, the document signature is invalid.

4. According to claim 2, a method for signing electronic documents based on fixed neural network steganography technology is characterized in that: Verifying the obtained signed document Doc′ specifically includes the following steps: The cloud platform publishes the signed document Doc′ to a public website; The client downloads the signed document Doc′ from the public website and transmits it to the cloud platform, and the client includes a signer and a verifier; The cloud platform extracts the stego-image C′ from the signed document Doc′, and uses the signed document Doc′ after removing the hidden image C′ as the verification document Doc″; The cloud platform uses the decoder in the fixed neural network steganography technology FNNS model to extract the verification value N of the extracted stego image C′, and returns the verification value N and the verification document Doc″ to the client; The client calculates the plaintext digest of the verification document Doc″ using the same hash function as the document signing step, and obtains the plaintext digest value H′ corresponding to the verification document Doc″; The client compares the document plaintext digest H in the verification value N with the digest value H′ corresponding to the verification document Doc″ to see if they are consistent. If they are consistent, the signature document Doc′ has not been tampered with. If they are inconsistent, the signature document Doc′ has been tampered with. When the signed document Doc′ has not been tampered with, the client generates a verification signature value M′ based on the signatory identity ID extracted from the verification value N and the document digest value H through the identification cryptographic signature algorithm; The client compares the signature value M in the verification value N with the verification signature value M′. If they are consistent, the document signature is valid; if they are inconsistent, the document signature is invalid.

5. According to claim 1, a method for signing electronic documents based on fixed neural network steganography technology is characterized in that: In the document signing step, the signer performs a digest operation on the document Doc to be signed through a hash function to generate a document plaintext digest H, which specifically includes the following steps: The signer performs a digest operation on the document Doc to be signed using the SM3 algorithm to generate a document plaintext digest H.

6. The electronic document signing method based on fixed neural network steganography technology according to claim 1 is characterized in that: In the document signing step, the cloud platform uses a fixed neural network steganography technology to write the verification value N into the signature image C to obtain a steganographic image C' containing the verification value, which specifically includes the following steps: The encoder in the fixed neural network steganography technology FNNS model is used to write the verification information N consisting of the signature value M, the document plaintext summary H, and the verified identity ID into the signature image.

7. The electronic document signing method based on fixed neural network steganography technology according to claim 6 is characterized in that: The encoder in the fixed neural network steganography technology FNNS model is used to write the signature value M, the document plaintext summary H, and the verification information N composed of the verified identity ID into the signature image, which is achieved by optimizing the following minimum binary cross entropy loss function: Among them, L BCE represents the binary cross entropy loss function, C' is the stego-image containing verification information, N is the binary information consisting of the signature value M to be embedded, the document plaintext summary H, and the verified identity ID, and F is the preset fixed decoder network SteganoGAN.

8. The electronic document signing method based on fixed neural network steganography technology according to claim 7 is characterized in that: The constraints for minimizing the binary cross entropy loss function are: ||C-C'|| ∞ ≤δ Where δ is the threshold of image change caused by the embedding process.

Images