Secure encryption and authentication method of face images based on PRO-FaceS

Abstract

The present invention relates to the field of face recognition technology, and discloses a method for secure encryption and authentication of face images based on PRO-FaceS, comprising the steps of: a terminal device calculates a perceived hash function value of an original face image, generates a key, encrypts the original face image based on a PRO-FaceS face encryption model and a key, and generates a protected face image; the terminal device stores the perceived hash function value and the protected face image to a server; a user uploads a face image to be authenticated to the terminal device, the terminal device calculates the perceived hash function value and transmits it to the server, the server locates the corresponding protected face image according to the perceived hash function value, and returns the protected face image to the terminal device; the terminal device uses a key to perform image restoration processing on the protected face image, calculates the Euclidean distance between the restored face image and the original face image, and obtains an authentication result. The present invention combines security and reversibility methods to protect facial privacy.

Description

Technical Field

[0001] The present invention relates to the field of face recognition technology, and in particular to a face image security encryption and authentication method based on PRO-FaceS. Background Art

[0002] With the continuous advancement of facial recognition technology, the convenience of public life has been greatly improved. However, the negative impact of this technology has also caused people to pay increasing attention to facial privacy protection. The ubiquitous collection of facial data has led to the widespread dissemination of personal facial information, which not only increases the risk of personal information leakage, but also leads to data abuse, posing a serious threat to personal privacy and security.

[0003] In the past few decades, facial privacy protection technologies (such as face anonymization, de-identification or de-ID technology) have been intensively studied to defend against privacy threats posed by unauthorized recognition systems. Early facial privacy protection technologies mainly used traditional blurring methods to blur facial features in images, and achieved facial image obfuscation through filter-based obfuscation techniques, image occlusion methods, and various image transformation techniques. These methods are effective in visual privacy protection and simple to operate, but have limitations in maintaining image naturalness and are susceptible to image reconstruction attacks. With the continuous advancement of deep learning technology, especially the rise of deep generation models such as generative adversarial networks (GANs), researchers have begun to explore more natural and practical facial privacy protection generation methods, but due to the application of crude GAN mechanisms, the visual quality is usually unsatisfactory. In recent years, image encryption methods have been widely studied to cope with scenarios where original images need to be forensically obtained. However, due to the diversity of faces to be protected, a single iteration or transformation model cannot produce different unique identities. Using preprocessed templates as the target of the privacy protection process allows the protected image to closely reflect the predefined template. Although this method performs well in terms of naturalness and reversibility of images, the generative model used is computationally inefficient and rarely used. Therefore, traditional blur-based methods still need to be improved in scenarios where facial privacy is protected. Summary of the invention

[0004] The purpose of the present invention is to protect facial privacy by combining security, reversibility and practicality, and to provide a facial image security encryption and authentication method based on PRO-FaceS.

[0005] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions:

[0006] The face image security encryption and authentication method based on PRO-FaceS includes the following steps:

[0007] Step 1, user registration phase: the user uploads the original face image to be registered to the terminal device through the client, the terminal device calculates the perceptual hash function value of the original face image and generates a key, and then encrypts the original face image based on the PRO-FaceS face encryption model and the generated key to generate a protected face image; finally, the terminal device transmits the calculated perceptual hash function value and the generated protected face image to the server for storage;

[0008] Step 2, user identity authentication stage: the user uploads the face image to be authenticated to the terminal device through the client, the terminal device calculates the perceived hash function value of the face image and transmits it to the server, the server locates the corresponding protected face image in the database according to the perceived hash function value, and returns the protected face image to the terminal device; the terminal device uses the key to perform image restoration on the protected face image, and then calculates the Euclidean distance between the restored face image and the original face image. If the value of the Euclidean distance is less than the set threshold, the user identity authentication is successful and the user identity is confirmed, otherwise the user identity authentication fails and the user identity is not confirmed.

[0009] Compared with the prior art, the present invention has the following beneficial effects:

[0010] (1) During the user registration phase, the terminal device will forward encrypt the original facial image to ensure the privacy of the user. At the same time, it will also generate a perceptual hash function value for the authenticated facial image, which is an irreversible feature representation that can effectively capture the unique features of the face. The processed protected facial image and the corresponding perceptual hash function value will be securely stored in the server's database. Even if the information in the database is attacked and data is leaked, the facial identity information stored therein will not be maliciously stolen. Since the original facial image has been encrypted and the perceptual hash function value itself does not contain any reversible personal information, attackers cannot recover recognizable facial data from it. In addition, the use of perceptual hashing technology means that even if there are slight changes in the image, the generated perceptual hash function value can still remain consistent, thereby improving the reliability and security of the system for facial recognition.

[0011] (2) During the identity authentication stage, the terminal device will calculate the perceived hash function value of the user's facial image that needs to be authenticated. The perceived hash function value generated by this process uniquely represents the user's facial features and provides a basis for subsequent matching. The terminal device uses the perceived hash function value to search for the corresponding encrypted protected facial image in the server. This step ensures that the system only focuses on protected data, thereby enhancing security. After finding the corresponding protected facial image, the server returns it to the terminal device, and the terminal device decrypts the protected facial image to restore the original facial image. During the recovery process, the correct key must be provided for successful decryption. If the wrong key is used, the decrypted image will be completely different from the original facial image. This design effectively prevents malicious users from attempting to restore the user's facial information through the wrong key. After the decryption is completed, the system will compare the Euclidean distance between the restored facial image and the original facial image. If the calculated result is less than the preset threshold, the authentication is successful and the user's identity is confirmed; otherwise, the authentication fails.

[0012] (3) Since the PRO-FaceS face encryption model uses the key K to participate in the encryption process, and the key K only exists in the terminal device, only the correct key can correctly restore the face image, ensuring that even if the data is leaked, personal privacy will not be exposed without the correct key; the PRO-FaceS face encryption model not only improves the security of face image storage, but also enhances the reliability of the authentication process. It is a method that combines security, reversibility and practicality, and provides a new solution for facial privacy protection.

[0013] (4) For model encryption, the PRO-FaceS face encryption model is a reversible image blurring technology that protects facial privacy through a secure process. Its core architecture is a reversible protection network composed of secure affine coupling block (SACB) streams. In the encryption process of this model, the input original face image is first pre-blurred to increase the blurriness of the image and improve the privacy protection effect. Next, the original face image and the pre-blurred image will be subjected to discrete wavelet transform (DWT) to decompose them into multiple wavelet subbands, which contain different frequency information of the original face image. To ensure the security of the encryption process, the key K generated by the terminal device will first use a key derivation function (PBKDF) to derive a binary code of a fixed length of W`×H` from the initial key k. After geometric transformation, the binary code is converted into a 2D binary map k` with the same dimension as the single channel of the input original face image. k` will be converted into a 2D binary map k` through a discrete The wavelet transform is converted into a four-channel key K; these wavelet subbands and key K are sequentially input into N consecutive secure affine coupling blocks (SACB) for processing to enhance the security of the image; after being processed by N secure affine coupling blocks (SACB), the wavelet subband of the last secure affine coupling block (SACB) is converted back to the spatial domain, and finally two output images are generated: a protected face image and a protected image byproduct, of which the protected image byproduct will be discarded to avoid unnecessary information leakage; during the recovery process, the system will use the key K to replace the lost protected image byproduct as the input of the recovery process, and process it in combination with the protected face image; when the correct key K is provided, the restored face image will be very similar to the original face image, ensuring user privacy while achieving effective identity recognition; if the wrong key is used, a significantly different image will be generated, which can effectively prevent malicious recovery and further ensure the security of user facial information. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0015] Figure 1 is a flow chart of the method of the present invention;

[0016] Figure 2 A schematic diagram of the process of user registration in an embodiment of the present invention;

[0017] Figure 3 A schematic diagram of the process of user identity authentication in an embodiment of the present invention;

[0018] Figure 4 Schematic diagram of the encryption and decryption process of the PRO-FaceS face encryption model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0020] It should be noted that similar numbers and letters represent similar items in the following figures, so once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance, or implying any such actual relationship or order between these entities or operations. In addition, the terms "connected", "connected", etc. can be directly connected between elements, or indirectly connected via other elements.

[0021] Embodiment 1:

[0022] The present invention is achieved through the following technical solutions: Figure 1 As shown, a method for secure encryption and authentication of facial images based on PRO-FaceS is implemented based on a facial authentication system, and the facial authentication system includes a client, a terminal device, and a server. The method includes the following steps:

[0023] Step 1, user registration stage: The user registration stage mainly realizes the encryption and storage of facial images. User i first uploads the original facial image Pi to be registered to the terminal device through the client. The terminal device calculates the perceptual hash function value hi of the original facial image Pi and generates a key K. Then, based on the PRO-FaceS facial encryption model and the generated key K, the original facial image Pi is encrypted to generate a protected facial image PPi, thereby ensuring the privacy and security of facial image data during transmission and storage; finally, the terminal device transmits the calculated perceptual hash function value hi and the generated protected facial image PPi to the server for storage; i represents the i-th user.

[0024] For more details, see Figure 2 , the step 1 specifically comprises the following steps:

[0025] In step 1-1, user i collects his original face image Pi through the client, and the client uploads the collected original face image Pi to the terminal device.

[0026] Step 1-2, the terminal device calculates the perceptual hash function value hi of the original face image Pi.

[0027] Step 1-3: The terminal device inputs the original face image Pi into the configured PRO-FaceS face encryption model and Perform pre-blurring processing and use any image blurring processor O to obtain the pre-blurred face image Y:

[0028]

[0029] Among them, W and H represent the width and height of the original face image Pi respectively.

[0030] In terms of vision, the pre-blurred face image Y should be significantly different from the original face image Pi, but the difference should not exceed the set threshold T:

[0031]

[0032] in, It represents the perceptual similarity index between the original face image Pi and the pre-blurred face image Y, which can be PSNR, SSIM or LPIPS, etc. The threshold T is set to constrain the visual privacy of the original face image Pi to be effectively preserved in the pre-blurred face image Y. Generally speaking, the image blur processor O can be a facial blurr of various forms and strengths, which is carried on the PRO-FaceS face encryption model.

[0033] In steps 1-4, the terminal device generates the key K required for the encryption process through the configured KeyGen algorithm, which is used to control the pseudo-random information of the recovery process.

[0034] The steps 1-4 specifically include the following steps:

[0035] Step 1-4-1, the terminal device generates an initial key k with arbitrary characters and length, and then derives a binary code with a fixed width W' and a fixed height H' from the initial key k based on a key derivation function (PBKDF);

[0036] Step 1-4-2, geometrically transform the binary code into a 2D binary mapping key k` of the same dimension as the single channel of the original face image Pi:

[0037]

[0038] Wherein, PBKDF(k) indicates the derivation of the initial key k based on the key derivation function; t(.) indicates the geometric transformation;

[0039] Step 1-4-3, convert the 2D binary key k` into a four-channel key K:

[0040]

[0041] Among them, DWT(.) represents the discrete wavelet transform operation; K is the key finally generated by the terminal device, and the dimension of K is (4, W / 2, H / 2). The key K is used to control the protection and restoration process of the face image.

[0042] In one practicable manner, key generation is implemented using PBKDF version 2 (i.e., PBKDF2) of a hash-based message authentication code (HMAC) pseudo-random function. The properties of PBKDF2 ensure that the KeyGen algorithm outputs pseudo-random information, and any bit difference in the initial key k will result in completely different and unrelated key mappings.

[0043] Step 1-5, the original face image Pi and the pre-blurred face image Y are processed by discrete wavelet transform (DWT) to generate wavelet subbands:

[0044]

[0045]

[0046] Among them, Pi D Represents the wavelet subband of the original face image Pi; Y D Represents the wavelet subband of the pre-blurred face image Y.

[0047] Step 1-6: Sub-band Pi of the original face image Pi D and the wavelet subband Y of the pre-blurred face image Y D Together with the key K, it is passed into N consecutive secure affine coupling blocks (SACB) for encryption processing to generate the protected image byproduct Q and the protected face image PPi:

[0048]

[0049] Where P(.) represents the encryption process of the secure affine coupling block. The generated protected image byproduct Q is discarded.

[0050] From a visual perspective, the protected face image PPi should be closer to the pre-blurred face image Y:

[0051]

[0052] in, Represents the perceptual similarity index between the protected face image PPi and the pre-blurred face image Y.

[0053] See also Figure 4 , the steps 1-6 specifically include the following steps:

[0054] Step 1-6-1, transform the original face image Pi's wavelet subband Pi D , wavelet subband Y of pre-blurred face image Y D The key K is passed into the first secure affine coupling block (SACB) for encryption to obtain the first original face image Pi 1+D , the first pre-blurred face image Y 1+D ;

[0055] Step 1-6-2: the first original face image Pi 1+D , the first pre-blurred face image Y 1+D The key K is passed into the second secure affine coupling block (SACB) for encryption to obtain the second original face image Pi 2+D , the second pre-blurred face image Y 2+D ;

[0056] Step 1-6-3, and so on, until the N-1th original face image Pi is N-1+D , N-1th pre-blurred face image Y N-1+D The key K is passed into the Nth secure affine coupling block (SACB) for encryption to obtain the Nth original face image Pi N+D , Nth pre-blurred face image Y N+D ;

[0057] Step 1-6-4, transform the Nth original face image Pi by inverse wavelet transform N+D Converting back to the spatial domain, we get the protected image byproduct Q:

[0058]

[0059] The protected image byproduct Q is discarded; the Nth pre-blurred face image Y is transformed by inverse wavelet transform N+D Convert back to the spatial domain and get the protected face image PPi:

[0060]

[0061] Wherein, IWT(.) represents the inverse wavelet transform operation.

[0062] In the steps 1-6-1 to 1-6-4, the decomposed wavelet subbands are further input into N SACBs arranged in sequence. These SACBs are responsible for performing a series of complex nonlinear transformations. Each SACB contains four independent nonlinear mapping functions: , each nonlinear mapping function performs a unique transformation operation on the input data during its working process:

[0063]

[0064] Among them, C=3 represents the number of RGB channels.

[0065] In each SACB, for the jth original face image Pi j+D and the jth pre-blurred face image Y j+D The conversion formula is:

[0066]

[0067]

[0068] Among them, || represents channel-level connection; ; a represents the product of the Sigmoid function and a constant factor. a not only acts as an activation function, but also effectively constrains the processed feature map to ensure that its value is within a reasonable range. After N ordered SACBs, the wavelet subband of the Nth SACB is converted back to the spatial domain, and finally the protected image byproduct Q and the protected face image PPi are output.

[0069] Step 1-7, the generated protected face image PPi and the perceived hash function value hi are transmitted to the server for storage.

[0070] In this solution, the core of the PRO-FaceS face encryption model is a reversible protection network composed of N secure affine coupling blocks (SACBs), which are encryption operations during forward processing and decryption operations during reverse processing. This solution improves the PRO-FaceS face encryption model by adding a new security mechanism, which embeds the key K as an encryption condition into each SACB, thereby improving the overall security of the network. In addition, the PRO-FaceS face encryption model also combines discrete wavelet transform (DWT) and inverse wavelet transform (IWT) technologies, which can not only effectively improve the performance of data hiding, but also greatly improve the processing speed. Through these technical improvements, the PRO-FaceS face encryption model can provide stronger privacy protection and data security while maintaining efficient computing.

[0071] Step 2, user identity authentication stage: The identity authentication stage mainly implements user identity authentication. The terminal device verifies whether the user's identity is legitimate by restoring the protected face image PPi. Similar to the user registration stage, user i first uploads the face image Fi to be authenticated to the terminal device through the client. The terminal device calculates the perceived hash function value fi of the face image Fi and transmits it to the server. The server locates the corresponding protected face image PPi in the database according to the perceived hash function value fi, and returns the protected face image PPi to the terminal device; the terminal device uses the key K to perform image restoration processing on the protected face image PPi, and then calculates the Euclidean distance between the restored face image FFi and the original face image Pi. If the value of the Euclidean distance is less than the set threshold TT, the user identity authentication is successful and the user identity is confirmed. Otherwise, the user identity authentication fails and the user identity is not confirmed.

[0072] For more details, see Figure 3 , the step 2 specifically comprises the following steps:

[0073] In step 2-1, user i collects a face image Fi that needs to be authenticated through the client, and the client uploads the collected face image Fi that needs to be authenticated to the terminal device.

[0074] In step 2-2, the terminal device calculates the perceived hash function value fi of the face image Fi that needs to be authenticated, and uploads the perceived hash function value fi to the server; the server searches for the corresponding protected face image PPi based on the perceived hash function value fi, and returns the protected face image PPi to the terminal device for image restoration processing.

[0075] Step 2-3: The terminal device generates a key K and uses the key K to generate a substitute image byproduct X N+1 , the key K is repeated three times in the channel direction to align with the dimensions of the wavelet subbands of the RGB image: X N+1 = K||K||K, is considered as an auxiliary input to the backward recovery process; using the surrogate image byproduct X N+1 To replace the protected image byproduct Q that has been discarded during the recovery process.

[0076] Step 2-4, the terminal device processes the protected face image PPi sent by the server through discrete wavelet transform (DWT) to generate wavelet subbands:

[0077]

[0078] Among them, PPi N+1 Represents the wavelet subband that protects the face image PPi.

[0079] Step 2-5, replace the image byproduct X N+1 , wavelet subband PPi of protected face image PPi N+1 and the key K are input into N consecutive secure affine coupled inversion blocks (SACB -1 ) is decrypted to generate the restored image byproduct Q` and the correct or incorrect restored face image FFi:

[0080]

[0081] Among them, P -1 (.) represents the decryption process of the secure affine coupled inversion block. The generated restored image byproduct Q` is discarded.

[0082] From a visual perspective, the generated correct or incorrect restored face image FFi should be closer to the original face image Pi:

[0083]

[0084] in, Indicates the perceptual similarity index between the correct or incorrect restored face image FFi and the original face image Pi.

[0085] See also Figure 4 , the steps 2-5 specifically include the following steps:

[0086] Step 2-5-1, replace the image byproduct X N+1 and the wavelet subband PPi of the protected face image PPi N+1 Together with the Nth secure affine coupled inversion block (SACB -1 ) to obtain the Nth image byproduct XN , Nth protected face image PPi N ;

[0087] Step 2-5-2: convert the Nth image byproduct X N , Nth protected face image PPi N Together with the N-1th secure affine coupled inversion block (SACB -1 ) to obtain the N-1th image byproduct X N-1 , N-1th protected face image PPi N -1 ;

[0088] Step 2-5-3, and so on, until the second image byproduct X is 2 , the second protected face image PPi 2 Together with the first secure affine coupling inversion block (SACB -1 ) to obtain the first image byproduct X 1 , the first protected face image PPi 1 ;

[0089] Step 2-5-4, transform the first image byproduct X through inverse wavelet transform 1 Convert back to the spatial domain and get the restored image byproduct Q`:

[0090]

[0091] The first protected face image PPi is transformed by inverse wavelet transform 1 Convert back to the spatial domain and get the correct or incorrect restored face image FFi:

[0092]

[0093] Wherein, IWT(.) represents the inverse wavelet transform operation.

[0094] In steps 2-5-1 to 2-5-4, X N+1 and PPi N+1 The N inverted SACBs are processed by the inverse double function based on the equation, and the key K is integrated in the same way as the encryption process. Each SACB -1 For the j-th image byproduct X j , the jth protected face image PPi j The conversion formula is:

[0095]

[0096]

[0097] After N ordered SACBs -1 After that, the first SACB -1 The wavelet subbands are reconverted back to the spatial domain, and finally the restored image byproduct Q` and the correct or incorrect restored face image FFi are output.

[0098] In step 2-3, the terminal device generates a substitute image byproduct X using the key K. N+1 If the key K used is correct, the restored face image FFi generated in the end will be very close to the original face image Pi, and the correct restored face image FFi will be generated; if the key K used is wrong, the restored face image FFi generated in the end will deviate greatly from the original face image Pi, and the wrong restored face image FFi will be generated. Assume is the wrong key, To restore the wrong face image, the following constraints are required:

[0099]

[0100] This face image recovery mechanism controlled by the key K reflects the security of the architecture of this solution.

[0101] Step 2-6, calculate the Euclidean distance between the correct or incorrect restored face image FFi and the original face image Pi. If the Euclidean distance is less than the set threshold TT, the user identity authentication is successful and the user identity is determined. Otherwise, the user identity authentication fails and the user identity is not confirmed.

[0102] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A facial image security encryption and authentication method based on PRO-FaceS is implemented based on a facial authentication system, which includes a client, a terminal device, and a server, and is characterized in that: The method comprises the following steps: Step 1, user registration phase: the user uploads the original face image to be registered to the terminal device through the client, the terminal device calculates the perceptual hash function value of the original face image and generates a key, and then encrypts the original face image based on the PRO-FaceS face encryption model and the generated key to generate a protected face image; finally, the terminal device transmits the calculated perceptual hash function value and the generated protected face image to the server for storage; The step 1 specifically comprises the following steps: Step 1-1, user i collects his original face image Pi through the client, and the client uploads the collected original face image Pi to the terminal device; Step 1-2, the terminal device calculates the perceptual hash function value hi of the original face image Pi; Step 1-3: The terminal device inputs the original face image Pi into the configured PRO-FaceS face encryption model, and performs an encryption on the original face image Pi∈R 3×W×H Perform pre-blur processing and use image blur processor O to obtain the pre-blurred face image Y: Y=O(Pi)∈R 3×W×H ; Where W and H represent the width and height of the original face image Pi respectively; Step 1-4, the terminal device generates the key K required for the encryption process through the configured KeyGen algorithm; Step 1-5, the original face image Pi and the pre-blurred face image Y are processed by discrete wavelet transform to generate wavelet subbands: Pi D =DWT(Pi); Y D =DWT(Y); Among them, Pi D Represents the wavelet subband of the original face image Pi; Y D Represents the wavelet subband of the pre-blurred face image Y; Step 1-6: Sub-band Pi of the original face image Pi D and the wavelet subband Y of the pre-blurred face image Y D Together with the key K, it is passed into N consecutive secure affine coupling blocks for encryption processing to generate the protected image byproduct Q and the protected face image PPi: PPi=P(Pi,Y,K)∈R 3×W×H ; Wherein, P(.) represents the encryption processing of the secure affine coupling block; The generated protection image byproduct Q is discarded; Step 1-7, the generated protected face image PPi and the perceived hash function value hi are transmitted to the server for storage; The steps 1-6 specifically include the following steps: Step 1-6-1, transform the original face image Pi's wavelet subband Pi D , wavelet subband Y of the pre-blurred face image Y D The key K is passed into the first secure affine coupling block for encryption to obtain the first original face image Pi 1+D , the first pre-blurred face image Y 1+D ; Step 1-6-2: the first original face image Pi 1+D , the first pre-blurred face image Y 1+D The key K is passed into the second secure affine coupling block for encryption to obtain the second original face image Pi 2+D , the second pre-blurred face image Y 2 +D ; Step 1-6-3, and so on, until the N-1th original face image Pi is N-1+D , N-1th pre-blurred face image Y N-1+D The key K is passed into the Nth secure affine coupling block for encryption, and the Nth original face image Pi is obtained. N+D , Nth pre-blurred face image Y N+D ; Step 1-6-4, transform the Nth original face image Pi by inverse wavelet transform N+D Converting back to the spatial domain, we get the protected image byproduct Q: Q=IWT(Pi N+D ); The protection image byproduct Q is discarded; The Nth pre-blurred face image Y is transformed by inverse wavelet transform N+D Convert back to the spatial domain and get the protected face image PPi: PPi=IWT(Y N+D ) Where IWT(.) represents the inverse wavelet transform operation; Step 2, user identity authentication stage: the user uploads the face image to be authenticated to the terminal device through the client, the terminal device calculates the perceived hash function value of the face image and transmits it to the server, the server locates the corresponding protected face image in the database according to the perceived hash function value, and returns the protected face image to the terminal device; the terminal device uses the key to perform image restoration processing on the protected face image, and then calculates the Euclidean distance between the restored face image and the original face image. If the value of the Euclidean distance is less than the set threshold, the user identity authentication is successful and the user identity is confirmed, otherwise the user identity authentication fails and the user identity is not confirmed; The step 2 specifically includes the following steps: Step 2-1, user i collects the face image Fi that needs to be authenticated through the client, and the client uploads the collected face image Fi that needs to be authenticated to the terminal device; Step 2-2, the terminal device calculates the perceived hash function value fi of the face image Fi to be authenticated, and uploads the perceived hash function value fi to the server; the server searches for the corresponding protected face image PPi according to the perceived hash function value fi, and returns the protected face image PPi to the terminal device for image restoration processing; Step 2-3: The terminal device generates a key K and uses the key K to generate a substitute image byproduct X N+1 :X N+1 = K||K||K; use alternative image byproduct X N+1 to replace the protected image byproduct Q that has been discarded during the recovery process; Step 2-4, the terminal device processes the protected face image PPi sent by the server through discrete wavelet transform to generate wavelet subbands: PPi N+1 =DWT(PPi); Among them, PPi N+1 Represents the wavelet subband that protects the face image PPi; Step 2-5, replace the image byproduct X N+1 , wavelet subband PPi of protected face image PPi N+1 The key K is input into N consecutive secure affine coupled inversion blocks for decryption processing, thereby generating a restored image byproduct Q` and a correct or incorrect restored face image FFi: FFi=P -1 (PPi N+1 ,X N+1 ); Among them, P -1 (.) represents the decryption processing of the secure affine coupled inversion block; The generated restored image byproduct Q` is discarded; Step 2-6, calculating the Euclidean distance between the correct or incorrect restored face image FFi and the original face image Pi. If the Euclidean distance is less than the set threshold TT, the user identity authentication is successful and the user identity is determined. Otherwise, the user identity authentication fails and the user identity is not confirmed. The steps 2-5 specifically include the following steps: Step 2-5-1, replace the image byproduct X N+1 and the wavelet subband PPi of the protected face image PPi N+1 The Nth image byproduct X is obtained by inputting them into the Nth secure affine coupled inversion block for decryption. N , Nth protected face image PPi N ; Step 2-5-2: convert the Nth image byproduct X N , Nth protected face image PPi N The N-1th secure affine coupled inversion block is decrypted to obtain the N-1th image byproduct X N-1 , N-1th protected face image PPi N-1 ; Step 2-5-3, and so on, until the second image byproduct X is 2 , the second protected face image PPi 2 The first image byproduct X is obtained by passing it into the first secure affine coupling inversion block for decryption. 1 , the first protected face image PPi 1 ; Step 2-5-4, transform the first image byproduct X through inverse wavelet transform 1 Convert back to the spatial domain and get the restored image byproduct Q`: Q`=IWT(X 1 ); The first protected face image PPi is transformed by inverse wavelet transform 1 Convert back to the spatial domain and get the correct or incorrect restored face image FFi: FFi=IWT(PPi 1 ); Wherein, IWT(.) represents the inverse wavelet transform operation.

2. The method for secure facial image encryption and authentication based on PRO-FaceS according to claim 1, characterized in that: The steps 1-4 specifically include the following steps: Step 1-4-1, the terminal device generates an initial key k with arbitrary characters and length, and derives a binary code with a fixed width W' and a fixed height H' from the initial key k based on a key derivation function; Step 1-4-2, geometrically transform the binary code into a 2D binary mapping key k` of the same dimension as the single channel of the original face image Pi: k`=t(PBKDF(k)),k∈{-1,1} W×H ; Wherein, PBKDF(k) indicates the derivation of the initial key k based on the key derivation function; t(.) indicates the geometric transformation; Step 1-4-3, convert the 2D binary key k` into a four-channel key K: K = DWT(k`); Wherein, DWT(.) represents discrete wavelet transform operation; K is the key finally generated by the terminal device.

3. The method for secure facial image encryption and authentication based on PRO-FaceS according to claim 1, characterized in that: In the steps 1-6-1 to 1-6-4, the secure affine coupling block performs the processing of the j-th original face image Pi j+D and the jth pre-blurred face image Y j+D The conversion formula is: P j+1+D =P j+D ·exp(a(ρ(P j+D ||K)))+η(Y j+D ||K); Y j+1+D =Y j+D ·exp(a(ω(P j+D ||K))+φ(P j+D ||K); Among them, || represents the channel-level connection; j∈[1,N]; a represents the product of the Sigmoid function and a constant factor; ω( ), φ( ), ρ( ), η( ) all represent nonlinear mapping functions.

4. The method for secure encryption and authentication of facial images based on PRO-FaceS according to claim 1, characterized in that: In the steps 2-5-1 to 2-5-4, the secure affine coupling inversion block is applied to the j-th image byproduct X j , the jth protected face image PPi j The conversion formula is: X j =(X j+1 -η(PPi j+1 ||K))·exp(-a(ρ(PPi j+1 ||K))); PPi j =(P j+1 -φ(X j ||K))·exp(-a(ω(X j ||K))); Among them, || represents the channel-level connection; j∈[1,N]; a represents the product of the Sigmoid function and a constant factor; ω( ), φ( ), ρ( ), η( ) all represent nonlinear mapping functions.

Images