A Multi-Image Encryption Method Based on 3D Face Key and Multiplexed Digital Holography

Abstract

This invention discloses a multi-image encryption method based on 3D face keys and multiplexed digital holography. The encryption steps include: using 3D face high-order data acquisition technology, chaotic face structured light phase mask generation technology, grating modulation technology, multiplexed digital holographic coding technology, and watermark embedding technology based on discrete wavelet domain singular value decomposition, multiple plaintext information images can be encrypted into an amplitude-type holographic ciphertext and embedded into a host image; the decryption steps include: the user must first calculate the face similarity using a 3D face multilayer perceptron neural network. If the similarity exceeds a preset threshold, authentication is successful. Then, the final decryption result is generated using watermark image extraction technology based on discrete wavelet transform, image decryption technology based on the Cramer-Kroni relation, and spectral filtering technology; otherwise, authentication fails. This method has advantages such as a large key space, high sensitivity, strong robustness of the 3D face key, and low decryption crosstalk.

Description

Technical Field

[0001] This invention relates to the field of information security technology, specifically to a multi-image encryption method based on three-dimensional face keys and multiplexed digital holography. Background Technology

[0002] In recent years, to protect sensitive information in digital images from theft and tampering by malicious attackers, encryption techniques for multiple images have gradually become a research hotspot for scholars both domestically and internationally. Among numerous encryption techniques, optical image encryption technology has attracted widespread attention from research groups at home and abroad due to its characteristics of fast encryption speed, strong security, large key space, high parallelism, and ability to process large-capacity images. The most representative example is the Random Phase Coding (DRPE) technique based on a 4f optical system proposed by Refregier and Javidi in 1995. Subsequently, advanced optical technologies such as ghost imaging, computational holography, metasurfaces, and digital holography have been gradually introduced into optical multi-image encryption methods. For example, Wu et al. proposed a multi-image encryption method based on computational ghost imaging. Zhu et al. proposed a multi-image encryption method based on chaotic mapping and computational holography. Zhao et al. proposed a multi-image encryption method based on polarized multiplexed metasurfaces. More recently, Li et al. proposed a multi-image encryption method based on three-step phase-shifting digital holography.

[0003] However, the aforementioned multi-image optical encryption methods generally suffer from high key complexity and insufficient practicality, particularly the difficulty in remembering or carrying random mask keys. Furthermore, existing encryption mechanisms often fail to reliably bind keys to user identities, resulting in keys lacking identifiable identifiers. Once a key is lost or stolen, attackers can directly use it to illegally access the system and steal sensitive image data, seriously threatening the information security of the encryption system. Against this backdrop, introducing biometric keys, especially facial recognition keys, is considered an effective improvement path. Due to the inherent advantages of facial features—uniqueness, portability, and ease of forgetting—an intrinsic link between keys and user identities can be established, significantly improving both security and the convenience of key management. Therefore, various research groups have recently begun to explore incorporating facial recognition keys into optical image encryption methods. For example, Wang et al. proposed an image encryption method based on facial recognition keys and bitonic sequences. Taheri et al. proposed an image encryption method based on facial recognition keys and DRPE. Later, Verma et al. proposed an optical image encryption method based on facial recognition keys, Fourier truncation, and natural logarithms. Recently, Su et al. also proposed a color image encryption method based on face keys and chaotic mapping.

[0004] However, the face keys used in the aforementioned optical encryption methods are all two-dimensional face keys, which share the common problems of weak security and poor robustness. Specifically, because two-dimensional face keys are limited by pixel dimensions, they are easily intercepted and copied by malicious attackers, who can then impersonate legitimate users and perform unauthorized decryption. Furthermore, when a legitimate user's face key is subjected to slight rotation, expression, or occlusion attacks during acquisition, the features extracted by the security system will be severely deviated, leading to authentication failure and decryption termination, thus seriously affecting the decryption experience of legitimate users. To address these problems, this invention discloses a multi-image encryption method based on a three-dimensional face key and multiplexed holography, aiming to achieve multi-image encryption with strong encryption security, a large key space, easy ciphertext storage, and high robustness of the face key. Summary of the Invention

[0005] The purpose of this invention is to provide a multi-image encryption method based on 3D face keys and multiplexed holography. This invention utilizes high-order 3D face data acquisition technology, chaotic face structured light phase mask generation technology, grating modulation technology, multiplexed digital holographic coding technology, and watermark embedding technology based on Discrete Wavelet Domain Singular Value Decomposition (DWT-SVD) to encrypt multiple plaintext images into a single amplitude-type holographic ciphertext and embed it into a host image. It has advantages such as strong security, large key space, easy storage of ciphertext, and high robustness of the face key. To achieve the above-mentioned objective, the technical solution provided by this invention is as follows: A multi-image encryption method based on 3D face keys and multiplexed digital holography, characterized by including the following encryption steps:

[0006] S1: High-level 3D face data extraction and encryption key generation: First, the original 3D face image of the encrypted user is acquired using a 3D face recognition device. Subsequently, 3D face image processing technology and 3D face high-order data extraction technology based on 3D face multilayer perceptron neural network were used to obtain... Average hash value of point cloud data and high-order feature vectors of the face And store it in an encrypted database; secondly, for The chaotic structured light phase mask for the face required in encryption step S2 is generated using chaotic mapping technology, a chaotic matrix generation program, and a structured light phase mask generation algorithm. and

[0007] S2: Encryption of Light Wave Generation and Multiplexed Digital Holographic Encoding: First, divide the K plaintext images into two groups. , Subsequently, grating modulation technology was used to process the plaintext image. The encryption fusion result is obtained by performing fusion operations separately. and Secondly, utilize and right and The encrypted light wave is obtained by performing double random phase encoding operations respectively. and Finally, regarding the wave... and Perform multiplexing digital holographic encoding operations to obtain encrypted holograms. .

[0008] S3: Watermark Ciphertext Generation Based on DWT-SVD: First, the Discrete Wavelet Transform (DWT) is used to generate the watermark ciphertext from the host image. Perform a decomposition operation to obtain its low-frequency components. Horizontal high-frequency components Vertical high-frequency components High-frequency components in the diagonal direction :

[0009]

[0010] in This is represented as a two-dimensional discrete wavelet transform. Then, the low-frequency components are analyzed. Perform singular value decomposition (SVD) to obtain the decomposition result. , , :

[0011]

[0012] in This is represented as a 2vctic differential value decomposition operation. Next, the encryption user sets the encryption embedding factor A and... and Perform an image embedding operation to obtain the embedding result. :

[0013]

[0014] Immediately afterwards Then perform singular value decomposition to obtain Decomposition results , , :

[0015]

[0016] Next, utilize , , right Perform a refactoring operation to obtain the refactoring result. :

[0017]

[0018] in for The transpose of . Finally, for , , The watermarked image obtained by performing the inverse discrete wavelet transform is the final ciphertext. :

[0019]

[0020] in This is represented as a two-dimensional discrete wavelet inverse transform. It is worth noting that in this step... and It will be stored in an encrypted database.

[0021] The encryption process is now complete.

[0022] Decryption steps:

[0023] J1: 3D Face Authentication and Decryption Key Acquisition: First, a 3D face recognition device is used to acquire the original 3D face image of the user. Subsequently, high-order feature vectors of the user's face were obtained by using 3D face image processing technology and 3D face high-order data extraction technology based on 3D face multilayer perceptron neural network. And calculate the facial similarity between the encrypted user and the decrypted user. ,like Greater than the preset threshold If the authentication is successful, the encrypted database stored in the database will be output. , , The face conjugate chaotic structured light phase mask required for decryption step S3 is generated using a chaotic matrix generation program and a conjugate structured light phase mask generation algorithm. and Then proceed to the next decryption step; otherwise, authentication fails and decryption terminates.

[0024] J2: Generation of watermark decryption results based on DWT-SVD: First, the ciphertext is decrypted using Discrete Wavelet Transform (DWT). Perform a decomposition operation to obtain its low-frequency components. Horizontal high-frequency components Vertical high-frequency components High-frequency components in the diagonal direction :

[0025]

[0026] Subsequently Perform singular value decomposition to obtain Decomposition results:

[0027]

[0028] Then, using the encrypted database stored and Perform watermark decryption to generate a decrypted hologram. :

[0029]

[0030] in for The transpose of .

[0031] J3: Multi-image decryption based on Kramer–Kronig relations: First, holographic decoding technology based on Kramer–Kronig relations is used to decrypt the holograms. Perform a decoding operation to obtain the light wave of the decrypted object. and ; and then utilize and right and Perform dual random phase decoding operations separately to obtain the decryption and fusion result. and Finally, spectral filtering techniques were used to respectively... and Perform filtering operations to obtain the final decryption result. and .

[0032] The aforementioned multi-image encryption method based on 3D face keys and multiplexed holography, wherein the encryption step S1, 3D face high-order data extraction and encryption key generation, includes the following steps:

[0033] (i) Acquisition of high-level 3D face data: First, the original 3D face image of the encrypted user is acquired using a 3D face acquisition device. Then calculate Hash mean of all point cloud data Subsequently, on To obtain by performing standardized operations Standardization results :

[0034]

[0035] Where N is Total number of vertices for The mean. Next, we will... Perform downsampling operation to generate downsampling results Finally, Input is fed into a 3D face multilayer perceptron neural network to obtain High-order feature vector of face :

[0036]

[0037] in It is a three-dimensional face multilayer perceptron neural network consisting of a 768-dimensional input layer, a 512-dimensional first hidden layer, a 256-dimensional second hidden layer, and a 128-dimensional output layer.

[0038] (ii) Generation of chaotic structured light phase mask for face: First, the chaotic structured light phase mask for face is generated... Perform chaotic mapping operations to generate chaotic initial values. :

[0039]

[0040] in Subsequently As the initial value of chaos, it is input into the chaotic system, where the general term of the chaotic sequence is... for:

[0041]

[0042] in Next, the chaotic sequence is arranged into a two-dimensional matrix to generate the face chaotic matrix. Next, the encryption user uses a simple set of digital keys and Generate the chaotic face structured light phase mask required for encryption step S2 :

[0043]

[0044] in and These are represented as Fresnel zone plates and radial Hilbert masks, respectively. The wavelength of light. Let be the radius of the Fresnel zone plate. The focal length of the Fresnel zone plate. Let the topological charge number be the radial Hilbert mask. For the spatial azimuth of the radial Hilbert mask, The imaginary unit, For phase operation, and .

[0045] The aforementioned multi-image encryption method based on three-dimensional face keys and multiplexed holography, wherein the encryption step S2, encryption object light wave generation and multiplexed digital holographic encoding, includes the following steps:

[0046] (i) Encrypted fusion image generation based on grating modulation: First, divide the K plaintext images into two groups. , Subsequently, grating modulation technology was used to respectively... , Perform image fusion operations to obtain the fused result. and :

[0047]

[0048] in and , The grating coefficients are expressed mathematically as follows:

[0049]

[0050] in, For the grating spatial frequency, and These are the angles between the grating line direction and the horizontal and vertical directions, respectively. and These are the length and width of the grating, respectively. .

[0051] (ii) Generation of encrypted light waves based on Fresnel diffraction domain double random phase coding: First, the encrypted fusion result is processed... and Perform one wavelength each time The diffraction distance is Fresnel diffraction was then performed, and the transformation result was compared with the first chaotic face structured light phase mask. and the third chaotic face structured light phase mask Multiply, and then perform another operation on the multiplication result with wavelength. The diffraction distance is The Fresnel diffraction was then performed, and the transformation result was compared with the second chaotic face structured light phase mask. and the fourth chaotic face structured light phase mask Finally, perform another multiplication on the result of this multiplication with a wavelength of... The diffraction distance is Fresnel diffraction thus yields the encrypted light wave. and :

[0052]

[0053] in Represented as wavelength The diffraction distance is Fresnel diffraction.

[0054] (iii) Encrypted hologram generation based on multiplexed digital holographic coding: introducing light waves related to the object. and Reference light waves of different axes and make , , Interference occurs on the holographic recording plane to generate an encrypted hologram. :

[0055]

[0056] in , and The angles between the directions of the principal rays are respectively and .

[0057] The aforementioned multi-image encryption method based on 3D face keys and multiplexed holography, wherein the encryption step J1, 3D face authentication and decryption key acquisition, includes the following steps:

[0058] (i) 3D face authentication: First, a 3D face acquisition device is used to acquire and decrypt the user's original 3D face image. Then, in turn, Perform the same normalization and downsampling operations as sub-step (ii) in encryption step S1 to generate The corresponding sampling results And input it into a 3D face multilayer perceptron neural network to obtain High-order feature vector of face Then calculate With encrypted database The cosine similarity is used as the facial similarity (FS) between the encryption and decryption users:

[0059]

[0060] in This represents a modulo operation. If Greater than the preset threshold If the authentication is successful, the encrypted database stored in the database will be output. , , , , , Conversely, if authentication fails, decryption will terminate.

[0061] (ii) Generation of face conjugate chaotic structured light phase mask: First, for Perform the same chaotic mapping and face chaotic matrix generation procedure as sub-step (ii) in encryption step S1 to generate the face chaotic matrix. Subsequently, decryption requires the user to input a simple digital key, and the face conjugate chaotic structured light phase mask required for encryption step S3 is generated using a face conjugate chaotic structured light phase mask generation program. and :

[0062]

[0063] in This is a conjugate operation.

[0064] The aforementioned multi-image encryption method based on 3D face keys and multiplexed holography, wherein the encryption step J3 is based on Kramer-Kronig relations for multi-image decryption, includes the following steps:

[0065] (i) Holographic decoding based on the Cramer-Kroni relation: First, construct an auxiliary complex amplitude according to the Cramer-Kroni relation. :

[0066]

[0067] in , and To decipher the object's light waves. Then... Perform exponentiation:

[0068]

[0069] in and These represent the real part extraction and imaginary part extraction operations, respectively. It can be represented as:

[0070]

[0071] Secondly, based on the Hilbert transform It can be represented as:

[0072]

[0073] in Cauchy principal value, As an intermediate variable, For coordinates in the spatial frequency domain, and These are the inverse Fourier transform and the inverse Fourier transform, respectively. This is a symbolic function. Next, we will utilize... and Perform a solution operation to obtain the decrypted object's light wave. and :

[0074]

[0075] in and They are respectively and The corresponding spatial filter.

[0076] (ii) Generation of decryption fusion results based on Fresnel diffraction domain double random phase decoding: First, the decryption fusion results are generated... and Perform one wavelength each time The diffraction distance is Fresnel inverse diffraction and the transformation result are respectively compared with , Multiply, and then perform the product again on wavelength. The diffraction distance is Fresnel inverse diffraction and respectively with , Multiply, and finally perform a separate operation on the product result with wavelength... The diffraction distance is The Fresnel inverse diffraction was used to obtain the decrypted fusion result. and

[0077]

[0078] in For wavelength The diffraction distance is Fresnel inverse diffraction.

[0079] (iii) Generation of decryption results based on spectral filtering: First, the decryption result is generated... and Perform a Fourier transform and filter the result using a spatial filter. Finally, perform an inverse Fourier transform on the filtered result to obtain the final decryption result. and :

[0080]

[0081] in and .

[0082] The beneficial effects of the present invention are as follows: (1) The present invention uses a chaotic matrix generation program, a radial Hilbert mask generation technique, and a phase modulation technique to generate a chaotic radial Hilbert mask used in the encryption step. The chaotic matrix generation program can provide multiple high-sensitivity digital keys, which can effectively increase the key space; (2) The present invention uses a multi-dimensional parameter selection hologram and a multi-dimensional multiplexing encryption hologram generation technique to encrypt multiple three-dimensional images into a phase-type holographic ciphertext; (3) The present invention uses a grating modulation technique, a multiplexed digital holographic coding technique, and a watermark embedding technique based on discrete wavelet domain singular value decomposition (DWT-SVD) to encrypt multiple plaintext information into an amplitude-type holographic ciphertext and embed it into a host image, which can effectively improve the storage and carrying capacity of the ciphertext; (4) The three-dimensional face key used in the present invention has strong security and the ability to resist cropping attacks and Gaussian noise attacks. Attached Figure Description

[0083] Figure 1 This is a flowchart of the encryption steps in a multi-image encryption method based on 3D face keys and multiplexed digital holography.

[0084] Figure 2 (a) and (b) respectively illustrate the generation of chaotic structured light phase masks and high-order facial feature vectors in a multi-image encryption method based on 3D face keys and multiplexed digital holography. Extraction flowchart.

[0085] Figure 3 The diagram illustrates the optical apparatus for sub-step (iii) in encryption step S2, where P is a polarizer, OL is an objective lens, PH is a pinhole, CL is a collimating lens, HWP is a half-wave plate, NPBS is a depolarizing beam splitter, BS is a beam splitter, M is a mirror, and L is a lens.

[0086] Figure 4 This is a flowchart of the final ciphertext generation process in a multi-image encryption method based on 3D face keys and multiplexed digital holography.

[0087] Figure 5 This is a flowchart of the decryption steps in a multi-image encryption method based on 3D face keys and multiplexed digital holography.

[0088] Figure 6This is a flowchart of the decryption and image reconstruction process in a multi-image encryption method based on 3D face keys and multiplexed digital holography.

[0089] Figure 7 (a1)-(a6) are plaintext images respectively. , , , , , (b1)-(b6) are the amplitude-type gratings corresponding to (a1)-(a6) respectively; (c1)-(c2) are the encryption and fusion results respectively. and (d1)-(d2) represent the spectral distributions corresponding to (c1)-(c2); (e1)-(e2) represent the original 3D face images of the encrypted user. and the corresponding downsampling results (f1)-(f4) are the chaotic structured light phase masks for human faces, respectively. (g1)-(g2) represent the encryption light waves, respectively. The amplitude and phase distributions; (h1)-(h2) are the encryption light waves respectively. The amplitude and phase distributions; (i1)-(i2) are the encrypted holograms respectively. and its spectral distribution; (j) is the host image. (k1)-(k2) are the values ​​generated in encryption step S3, respectively. and (l) is the watermark image, i.e., the final ciphertext. .

[0090] Figure 8 This is the decryption result of a multi-image encryption method based on 3D face keys and multiplexed digital holography when all keys are correct.

[0091] Figure 9 This is an analysis of the security and effectiveness of the 3D face key and digital key in a multi-image encryption method based on 3D face key and multiplexed digital holography.

[0092] Figure 10 This paper presents the analysis results of the robustness of the three-dimensional face key against rotation attacks, expression attacks, and occlusion attacks in a multi-image encryption method based on three-dimensional face keys and multiplexed digital holography.

[0093] Figure 11 (a)-(h) represent the first diffraction distance deviations, respectively. Second diffraction distance deviation Third diffraction distance deviation Wavelength key deviation First topological charge number key deviation Second topological charge key deviation First focal length key deviation Second focal length key deviation The corresponding sensitivity curve. Detailed Implementation

[0094] The present invention will be further described below with reference to the embodiments and accompanying drawings, but this should not be construed as limiting the present invention.

[0095] Example: A multi-image encryption method based on 3D face keys and multiplexed digital holography, including as follows Figure 1 The encryption steps shown are as follows: S1: 3D face high-order data extraction and encryption key generation: First, the original 3D face image of the encrypted user is acquired using a 3D face recognition device. Subsequently, 3D face image processing technology and 3D face high-order data extraction technology based on 3D face multilayer perceptron neural network were used to obtain... Average hash value of point cloud data and high-order feature vectors of the face And store it in an encrypted database; secondly, for The chaotic structured light phase mask for the face required in encryption step S2 is generated using chaotic mapping technology, a chaotic matrix generation program, and a structured light phase mask generation algorithm. and The step described above involves generating a chaotic structured light phase mask for the face and generating high-order feature vectors for the face. The extraction flowcharts are as follows: Figure 2 As shown in (a) and (b), the steps include:

[0096] (i) Acquisition of high-level 3D face data: First, the original 3D face image of the encrypted user is acquired using a 3D face acquisition device. Then calculate Decimal hash mean of all point cloud data Subsequently, on To obtain by performing standardized operations Standardization results :

[0097]

[0098] Where N is Total number of vertices for The mean. Next, we will... Perform downsampling operation to generate downsampling results Finally, Input is fed into a 3D face multilayer perceptron neural network to obtain High-order feature vector of face :

[0099]

[0100] in It is a three-dimensional face multilayer perceptron neural network consisting of a 768-dimensional input layer, a 512-dimensional first hidden layer, a 256-dimensional second hidden layer, and a 128-dimensional output layer.

[0101] (ii) Generation of chaotic structured light phase mask for face: First, the chaotic structured light phase mask for face is generated... Perform chaotic mapping operations to generate chaotic initial values. :

[0102]

[0103] in Subsequently As the initial value of chaos, it is input into the chaotic system, where the general term of the chaotic sequence is... for:

[0104]

[0105] in Next, the chaotic sequence is arranged into a two-dimensional matrix to generate the face chaotic matrix. Next, the encryption user uses a simple set of digital keys and Generate the chaotic face structured light phase mask required for encryption step S2 :

[0106]

[0107] in and These are represented as Fresnel zone plates and radial Hilbert masks, respectively. The wavelength of light. Let be the radius of the Fresnel zone plate. The focal length of the Fresnel zone plate. Let the topological charge number be the radial Hilbert mask. For the spatial azimuth of the radial Hilbert mask, The imaginary unit, For phase operation, and .

[0108] S2: Encryption of Light Wave Generation and Multiplexed Digital Holographic Encoding: First, divide the K plaintext images into two groups. , Subsequently, grating modulation technology was used to process the plaintext image. The encryption fusion result is obtained by performing fusion operations separately. and Secondly, utilize and right and The encrypted light wave is obtained by performing double random phase encoding operations respectively. and Finally, regarding the wave... and Perform multiplexing digital holographic encoding operations to obtain encrypted holograms. Its flowchart is as follows Figure 4 As shown, it includes the following steps:

[0109] (i) Encrypted fusion image generation based on grating modulation: First, divide the K plaintext images into two groups. , Subsequently, grating modulation technology was used to respectively... , Perform image fusion operations to obtain the fused result. and :

[0110]

[0111] in and , The grating coefficients are expressed mathematically as follows:

[0112]

[0113] in, For the grating spatial frequency, and These are the angles between the grating line direction and the horizontal and vertical directions, respectively. and These are the length and width of the grating, respectively. .

[0114] (ii) Generation of encrypted light waves based on Fresnel diffraction domain double random phase encoding (optical schematic diagram of this sub-step is shown in Figure 1) Figure 3 As shown): First, the encrypted fusion result... and Perform one wavelength each time The diffraction distance is Fresnel diffraction was then performed, and the transformation result was compared with the first chaotic face structured light phase mask. and the third chaotic face structured light phase mask Multiply, and then perform another operation on the multiplication result with wavelength. The diffraction distance is The Fresnel diffraction was then performed, and the transformation result was compared with the second chaotic face structured light phase mask. and the fourth chaotic face structured light phase mask Finally, perform another multiplication on the result of this multiplication with a wavelength of... The diffraction distance is Fresnel diffraction thus yields the encrypted light wave. and :

[0115]

[0116] in Represented as wavelength The diffraction distance is Fresnel diffraction.

[0117] (iii) Encrypted hologram generation based on multiplexed digital holographic coding: introducing light waves related to the object. and Reference light waves of different axes and make , , Interference occurs on the holographic recording plane to generate an encrypted hologram. :

[0118]

[0119] in , and The angles between the directions of the principal rays are respectively and .

[0120] S3: Watermark ciphertext generation based on DWT-SVD, where the watermark ciphertext generation flowchart is as follows: Figure 4 As shown: First, the host image is processed using Discrete Wavelet Transform (DWT). Perform a decomposition operation to obtain its low-frequency components. Horizontal high-frequency components Vertical high-frequency components High-frequency components in the diagonal direction :

[0121]

[0122] in This is represented as a two-dimensional discrete wavelet transform. Then, the low-frequency components are analyzed. Perform singular value decomposition (SVD) to obtain the decomposition result. , , :

[0123]

[0124] in This is represented as a 2vctic differential value decomposition operation. Next, the encryption user sets the encryption embedding factor A and... and Perform an image embedding operation to obtain the embedding result. :

[0125]

[0126] Immediately afterwards Then perform singular value decomposition to obtain Decomposition results , , :

[0127]

[0128] Next, utilize , , right Perform a refactoring operation to obtain the refactoring result. :

[0129]

[0130] in for The transpose of . Finally, for , , The watermarked image obtained by performing the inverse discrete wavelet transform is the final ciphertext. :

[0131]

[0132] in This is represented as a two-dimensional discrete wavelet inverse transform. It is worth noting that in this step... and It will be stored in an encrypted database.

[0133] This concludes the encryption process.

[0134] Next, according to Figure 5 The decryption steps shown are Figure 6The decryption flowchart shown executes the decryption process:

[0135] J1: 3D Face Authentication and Decryption Key Acquisition: First, a 3D face recognition device is used to acquire the original 3D face image of the user. Subsequently, high-order feature vectors of the user's face were obtained by using 3D face image processing technology and 3D face high-order data extraction technology based on 3D face multilayer perceptron neural network. And calculate the facial similarity between the encrypted user and the decrypted user. ,like Greater than the preset threshold If the authentication is successful, the encrypted database stored in the database will be output. , , The face conjugate chaotic structured light phase mask required for decryption step S3 is generated using a chaotic matrix generation program and a conjugate structured light phase mask generation algorithm. and Then proceed to the next decryption step; otherwise, authentication fails and decryption terminates. This includes the following steps:

[0136] (i) 3D face authentication: First, a 3D face acquisition device is used to acquire and decrypt the user's original 3D face image. Then, in turn, Perform the same normalization and downsampling operations as sub-step (ii) in encryption step S1 to generate The corresponding sampling results And input it into a 3D face multilayer perceptron neural network to obtain High-order feature vector of face Then calculate With encrypted database The cosine similarity is used as the facial similarity (FS) between the encryption and decryption users:

[0137]

[0138] in This represents a modulo operation. If Greater than the preset threshold If the authentication is successful, the encrypted database stored in the database will be output. , and Conversely, if authentication fails, decryption will terminate.

[0139] (ii) Generation of face conjugate chaotic structured light phase mask: First, for Perform the same chaotic mapping and face chaotic matrix generation procedure as sub-step (ii) in encryption step S1 to generate the face chaotic matrix. Subsequently, decryption requires the user to input a simple digital key, and the face conjugate chaotic structured light phase mask required for encryption step S3 is generated using a face conjugate chaotic structured light phase mask generation program. and :

[0140]

[0141] in This is a conjugate operation.

[0142] J2: Generation of watermark decryption results based on DWT-SVD: First, the ciphertext is decrypted using Discrete Wavelet Transform (DWT). Perform a decomposition operation to obtain its low-frequency components. Horizontal high-frequency components Vertical high-frequency components High-frequency components in the diagonal direction :

[0143]

[0144] Subsequently Perform singular value decomposition to obtain Decomposition results:

[0145]

[0146] Then, using the encrypted database stored and Perform watermark decryption to generate a decrypted hologram. :

[0147]

[0148] in for The transpose of .

[0149] J3: Multi-image decryption based on Kramer–Kronig relations: First, holographic decoding technology based on Kramer–Kronig relations is used to decrypt the holograms. Perform a decoding operation to obtain the light wave of the decrypted object. and ; and then utilize and right and Perform dual random phase decoding operations separately to obtain the decryption and fusion result. and Finally, spectral filtering techniques were used to respectively... and Perform filtering operations to obtain the final decryption result. and This includes the following steps:

[0150] (i) Holographic decoding based on the Cramer-Kroni relation: First, construct an auxiliary complex amplitude according to the Cramer-Kroni relation. :

[0151]

[0152] in , and To decipher the object's light waves. Then... Perform exponentiation:

[0153]

[0154] in and These represent the real part extraction and imaginary part extraction operations, respectively. It can be represented as:

[0155]

[0156] Secondly, based on the Hilbert transform It can be represented as:

[0157]

[0158] in Cauchy principal value, As an intermediate variable, For coordinates in the spatial frequency domain, and These are the inverse Fourier transform and the inverse Fourier transform, respectively. This is a symbolic function. Next, we will utilize... and Perform a solution operation to obtain the decrypted object's light wave. and :

[0159]

[0160] in and They are respectively and The corresponding spatial filter.

[0161] (ii) Generation of decryption fusion results based on Fresnel diffraction domain double random phase decoding: First, the decryption fusion results are generated... and Perform one wavelength each time The diffraction distance is Fresnel inverse diffraction and the transformation result are respectively compared with , Multiply, and then perform the product again on wavelength. The diffraction distance is Fresnel inverse diffraction and respectively with , Multiply, and finally perform a separate operation on the product result with wavelength... The diffraction distance is The Fresnel inverse diffraction was used to obtain the decrypted fusion result. and

[0162]

[0163] in For wavelength The diffraction distance is Fresnel inverse diffraction.

[0164] (iii) Generation of decryption results based on spectral filtering: First, the decryption result is generated... and Perform a Fourier transform and filter the result using a spatial filter. Finally, perform an inverse Fourier transform on the filtered result to obtain the final decryption result. and :

[0165]

[0166] in and .

[0167] The content of the present invention will be further explained below with reference to the accompanying drawings: First, select as follows Figure 7 The number of pixels shown in (a1)-(a6) is The grayscale images (“Barbara”, “Mandrill”, “Pepper”, “Pirates”, “Pirates”, “Earth”) are treated as plaintext images and divided into two groups. , Then utilize, such as Figure 7 The different gratings shown in (b1)-(b3) and (b4)-(b6) respectively... and Perform grating modulation operation to obtain such Figure 7 The encryption fusion results shown in (c1)-(c2) and .in and The spectral distributions are respectively as follows Figure 7 As shown in (d1)-(d2). Next, a 3D face acquisition device is used to acquire images as follows: Figure 7 (e1) shows the original 3D face image of the encrypted user. and calculate Decimal hash mean of all point cloud data ( ), then on Perform standardization and downsampling operations sequentially to obtain The corresponding downsampling results (like Figure 7 (as shown in e2) is input into a 3D face multilayer perceptron neural network to obtain... High-order feature vector of face The encrypted user then sets the focal length of the Fresnel wave plate. The topological charge numbers of the radial Hilbert masks are 30, 40, 50, and 60, respectively. The numbers are 3, 4, 5, and 6 respectively, and the following are generated using chaotic mapping technology, a face chaotic matrix generation program, and a structured light phase mask generation algorithm: Figure 7 The chaotic structured light phase mask for the face shown in (f1)-(f4) Subsequently, they respectively... and Perform one wavelength Diffraction distance Fresnel diffraction; then the results of this calculation were compared with... and Multiply; then perform wavelength multiplication again on the product result. Diffraction distance Fresnel diffraction, and the results of this calculation are respectively compared with and Multiply; finally, perform wavelength multiplication again on the product result. Diffraction distance Fresnel diffraction thus yields the amplitude and phase distributions as follows: Figure 7 The encryption light waves shown in (g1)-(g2) and 7(h1)-(h2) and Next, we will introduce the reference light. And by using multiplexed digital holographic coding technology to perform off-axis digital holographic coding operations, the amplitude and spectral distributions are obtained as follows: Figure 7 The encrypted holograms shown in (i1)-(i2) Finally, we introduce, for example... Figure 7 The host image (“Tree”) shown in (j) And using DWT-SVD watermarking technology to Embedded to Thus, the following is generated: Figure 7 The keys shown in (k1)-(k2) and And such as Figure 7 The final encrypted image shown in (l) .

[0168] The following is based on Figure 5 The decryption steps shown perform multi-image decryption, requiring the user's 3D face key and digital keys (wavelength key, three diffraction distance keys, four Fresnel zone plate focal length keys, and four topological charge number keys). When all keys are correct, and The decryption results are as follows: Figure 8 As shown in (a1)-(a3) and (b1)-(b3). Furthermore, to evaluate the decryption quality, the correlation coefficient (CC) is used to assess the degree of correlation between each decryption result and the original plaintext image, i.e.:

[0169]

[0170] in and They are respectively Original plaintext image and decryption result in grayscale value at that location. and They are respectively and The average pixel grayscale value. From Figure 8 It can be seen that when all keys are correct, the correlation coefficient between the original plaintext image (“Barbara”, “Mandrill”, “Pepper”, “Pirates”, “Pirates”, “Earth”) and the decryption result is greater than 0.945. Therefore, it can be proved that when all keys are correct, the original plaintext image can be reconstructed correctly with high quality.

[0171] The validity and security of the 3D face key and digital key in this invention will now be examined. The applicant randomly collected over 500 3D face data from different attackers as erroneous 3D face keys. Figure 9 The decryption results are given when both the 3D face key and the digital key are incorrect. It should be noted that... Figure 9 Only the topological charge number key is shown. , Focal key , Wavelength key Diffraction distance key , , Original plaintext image when an error occurs The corresponding decryption result, The corresponding decryption result and Figure 9 Similar. From Figure 9 It can be seen that when the digital key is incorrect, the decryption result exhibits significant noise distribution and the CC value between it and the original plaintext image is less than 0.02. However, when the 3D face key is incorrect, the face similarity (FS) between the decryption user and the encryption user is significantly lower than the preset threshold. This leads to authentication failure and ultimately, decryption termination. Therefore, the effectiveness and security of the digital key and the 3D face key in the proposed encryption method can be proven.

[0172] The robustness of the 3D face key in this invention is examined below, including robustness against rotation attacks, robustness against facial expression attacks, and robustness against occlusion attacks. Figure 10 Robustness analysis results of the 3D face key in this invention are presented. From Figure 10 It can be seen that even when the correct 3D face key is subjected to a rotation angle of... Even after a rotation attack, the face similarity (FS) between the face key and a 3D face key that has not been subjected to a rotation attack is still higher than a preset threshold. Thus, authentication is successful; when the correct 3D face key is subjected to different facial expression attacks, its similarity (FS) with the 3D face key that has not been subjected to facial expression attacks is still higher than a preset threshold. Thus, authentication is successful; even when the correct 3D face key is subjected to a blocking attack with a cropping ratio of less than 50%, its face similarity (FS) with the 3D face key that has not been subjected to the blocking attack is still higher than a preset threshold. Thus, authentication is successful. Therefore, it can be proven that the three-dimensional face key used in this invention has high robustness against rotation attacks, facial expression attacks, and occlusion attacks.

[0173] Finally, the sensitivity of the digital key in this invention is examined. It should be noted that... Figure 11 (a)-(h) represent the first diffraction distance deviations, respectively. Second diffraction distance deviation Third diffraction distance deviation Wavelength key deviation First topological charge number key deviation Second topological charge key deviation First focal length key deviation Second focal length key deviation The corresponding sensitivity curve reflects the decryption result when the key deviates. With the original plaintext The functional relationship between the average CC values. Key bias of the third topology load number. Fourth topological charge number key deviation Third focal length key deviation Fourth focal length key deviation The corresponding decryption result With the original plaintext The functional relationship between the average CC value and its key sensitivity curve is as follows: Figure 11 The curves shown in (e)-(h) have similar characteristics. From... Figure 11 As can be seen, when any digital key has a slight deviation, the average CC value between the decryption result and the original plaintext image drops sharply to close to 0. Only when all digital key deviations are 0 can the original plaintext image be correctly recovered, thus proving that the digital key in this invention has high sensitivity.

[0174] The above description is a specific illustration of the present invention, and not a limitation thereof. Those skilled in the art can make various equivalent technical solutions without departing from the scope of the present invention; therefore, all equivalent technical solutions should fall within the patent protection scope of the present invention.

Claims

1. A multi-image encryption method based on three-dimensional face keys and multiplexed digital holography, characterized in that: Including encryption steps: S1: High-level 3D face data extraction and encryption key generation: First, the original 3D face image of the encrypted user is acquired using a 3D face recognition device. ; Subsequently, 3D face image processing technology and high-order 3D face data extraction technology based on 3D face multilayer perceptron neural network were used to obtain... Average hash value of point cloud data and high-order feature vectors of the face And store it in an encrypted database; secondly, for The chaotic structured light phase mask for the face required in encryption step S2 is generated using chaotic mapping technology, a chaotic matrix generation program, and a structured light phase mask generation algorithm. and ; S2: Encryption of Light Wave Generation and Multiplexed Digital Holographic Encoding: First, divide the K plaintext images into two groups. , Subsequently, grating modulation technology was used to process the plaintext image. The encryption fusion result is obtained by performing fusion operations separately. and Secondly, utilize and right and The encrypted light wave is obtained by performing double random phase encoding operations respectively. and Finally, regarding the wave... and Perform multiplexing digital holographic encoding operations to obtain encrypted holograms. ; S3: Watermark Ciphertext Generation Based on DWT-SVD: First, the Discrete Wavelet Transform (DWT) is used to generate the watermark ciphertext from the host image. Perform a decomposition operation to obtain its low-frequency components. Horizontal high-frequency components Vertical high-frequency components High-frequency components in the diagonal direction : in This is represented as a two-dimensional discrete wavelet transform; subsequently, the low-frequency components are analyzed. Perform singular value decomposition (SVD) to obtain the decomposition result. , , : in This is represented as a 2vctic differential value decomposition operation; secondly, the encrypted user sets the encryption embedding factor A and performs... and Perform an image embedding operation to obtain the embedding result. : Immediately afterwards Then perform singular value decomposition to obtain Decomposition results , , : Next, utilize , , right Perform a refactoring operation to obtain the refactoring result. : in for The transpose; finally, for , , The watermarked image obtained by performing the inverse discrete wavelet transform is the final ciphertext. : in This is represented as a two-dimensional discrete wavelet inverse transform; it is worth noting that in this step... and It will be stored in an encrypted database; The encryption process is now complete. Decryption steps: J1: 3D Face Authentication and Decryption Key Acquisition: First, a 3D face recognition device is used to acquire the original 3D face image of the user. Subsequently, high-order feature vectors of the user's face were obtained by using 3D face image processing technology and 3D face high-order data extraction technology based on 3D face multilayer perceptron neural network. And calculate the facial similarity between the encrypted user and the decrypted user. ,like Greater than the preset threshold If the authentication is successful, the encrypted database stored in the database will be output. , , The face conjugate chaotic structured light phase mask required for decryption step S3 is generated using a chaotic matrix generation program and a conjugate structured light phase mask generation algorithm. and And proceed to the next decryption step; Conversely, if authentication fails, decryption will terminate. J2: Generation of watermark decryption results based on DWT-SVD: First, the ciphertext is decrypted using Discrete Wavelet Transform (DWT). Perform a decomposition operation to obtain its low-frequency components. Horizontal high-frequency components Vertical high-frequency components High-frequency components in the diagonal direction : Subsequently Perform singular value decomposition to obtain Decomposition results: Then, using the encrypted database stored... and Perform watermark decryption to generate a decrypted hologram. : in for transpose; J3: Multi-image decryption based on Kramer–Kronig relations: First, holographic decoding technology based on Kramer–Kronig relations is used to decrypt the holograms. Perform a decoding operation to obtain the light wave of the decrypted object. and ; and then utilize and right and Perform dual random phase decoding operations separately to obtain the decryption and fusion result. and Finally, spectral filtering techniques were used to respectively... and Perform filtering operations to obtain the final decryption result. and ; This concludes the decryption process.

2. The multi-image encryption method based on three-dimensional face key and multiplexed digital holography according to claim 1, characterized in that: The specific process of encryption step S1 is as follows: (i) Acquisition of high-level 3D face data: First, the original 3D face image of the encrypted user is acquired using a 3D face acquisition device. Then calculate Hash mean of all point cloud data ; then on To obtain by performing standardized operations Standardization results : Where N is Total number of vertices for The mean; next, for Perform downsampling operation to generate downsampling results Finally, Input is fed into a 3D face multilayer perceptron neural network to obtain High-order feature vector of face : in It is a three-dimensional face multilayer perceptron neural network consisting of a 768-dimensional input layer, a 512-dimensional first hidden layer, a 256-dimensional second hidden layer, and a 128-dimensional output layer. (ii) Generation of chaotic structured light phase mask for face: First, the chaotic structured light phase mask for face is generated... Perform chaotic mapping operations to generate chaotic initial values. : in ; then As the initial value of chaos, it is input into the chaotic system, where the general term of the chaotic sequence is... for: in Secondly, the chaotic sequence is arranged into a two-dimensional matrix to generate the face chaotic matrix. Next, the encrypted user uses a simple set of digital keys and Generate the chaotic face structured light phase mask required for encryption step S2 : in and These are represented as Fresnel zone plates and radial Hilbert masks, respectively. The wavelength of light. Let be the radius of the Fresnel zone plate. The focal length of the Fresnel zone plate. Let the topological charge number be the radial Hilbert mask. For the spatial azimuth of the radial Hilbert mask, The imaginary unit, For phase operation, and .

3. The multi-image encryption method based on three-dimensional face key and multiplexed digital holography according to claim 1, characterized in that: The specific process of encryption step S2 is as follows: (i) Encrypted fusion image generation based on grating modulation: First, divide the K plaintext images into two groups. , Subsequently, grating modulation technology was used to respectively... , Perform image fusion operations to obtain the fusion result. and : in and , The grating coefficients are expressed mathematically as follows: in, For the grating spatial frequency, and These are the angles between the grating line direction and the horizontal and vertical directions, respectively. and These are the length and width of the grating, respectively. ; (ii) Generation of encrypted light waves based on Fresnel diffraction domain double random phase coding: First, the encrypted fusion result is processed... and Perform one wavelength each time The diffraction distance is Fresnel diffraction was then performed, and the transformation result was compared with the first chaotic face structured light phase mask. and the third chaotic face structured light phase mask Multiply, and then perform another operation on the multiplication result with wavelength. The diffraction distance is The Fresnel diffraction was then performed, and the transformation result was subsequently compared with the second chaotic face structured light phase mask. and the fourth chaotic face structured light phase mask Finally, perform another multiplication on the result of this multiplication with a wavelength of... The diffraction distance is Fresnel diffraction thus yields the encrypted light wave. and : in Represented as wavelength The diffraction distance is Fresnel diffraction; (iii) Encrypted hologram generation based on multiplexed digital holographic coding: introducing light waves related to the object. and Reference light waves of different axes and make , , Interference occurs on the holographic recording plane to generate an encrypted hologram. : in , and The angles between the directions of the principal rays are respectively and .

4. The three-dimensional multi-image encryption method based on multidimensional multiplexed holograms according to claim 1, characterized in that: The specific process of decryption step J1 is as follows: (i) 3D face authentication: First, a 3D face acquisition device is used to acquire and decrypt the user's original 3D face image. Then, in turn, Perform the same normalization and downsampling operations as sub-step (ii) in encryption step S1 to generate The corresponding sampling results And input it into a 3D face multilayer perceptron neural network to obtain High-order feature vector of face ; then calculate With encrypted database The cosine similarity is used as the facial similarity (FS) between the encryption and decryption users: in This is represented as a modulo operation; if Greater than the preset threshold If the authentication is successful, the encrypted database stored in the database will be output. , , , , , Conversely, if authentication fails, decryption will terminate. (ii) Generation of face conjugate chaotic structured light phase mask: First, for Perform the same chaotic mapping and face chaotic matrix generation procedure as sub-step (ii) in encryption step S1 to generate the face chaotic matrix. Subsequently, decryption requires the user to input a simple digital key, and the face conjugate chaotic structured light phase mask required for encryption step S3 is generated using a face conjugate chaotic structured light phase mask generation program. and : in This is a conjugate operation.

5. A three-dimensional multi-image encryption method based on multidimensional multiplexed holograms according to claim 1, characterized in that: The specific process of decryption step J3 is as follows: (i) Holographic decoding based on the Cramer-Kroni relation: First, construct an auxiliary complex amplitude according to the Cramer-Kroni relation. : in , and To decipher the object's light waves; subsequently, Perform exponentiation: in and These are respectively represented as real part extraction and imaginary part extraction operations; where It can be represented as: Secondly, based on the Hilbert transform It can be represented as: in Cauchy principal value, As an intermediate variable, For coordinates in the spatial frequency domain, and These are the inverse Fourier transform and the inverse Fourier transform, respectively. For symbolic functions; next, we will use... and Perform a solution operation to obtain the decrypted object's light wave. and : in and They are respectively and The corresponding spatial filter; (ii) Generation of decryption fusion results based on Fresnel diffraction domain double random phase decoding: First, the decryption fusion results are generated... and Perform one wavelength each time The diffraction distance is Fresnel inverse diffraction and the transformation result are respectively compared with , Multiply, and then perform the product again on wavelength. The diffraction distance is Fresnel inverse diffraction and respectively with , Multiply, and finally perform a separate operation on the product result with wavelength... The diffraction distance is The Fresnel inverse diffraction was used to obtain the decrypted fusion result. and in For wavelength The diffraction distance is Fresnel inverse diffraction; (iii) Generation of decryption results based on spectral filtering: First, the decryption result is generated... and Perform a Fourier transform and filter the result using a spatial filter. Finally, perform an inverse Fourier transform on the filtered result to obtain the final decryption result. and : in and .

Images