A cloud data access control method based on image translation

Abstract

The present invention discloses a cloud data access control method based on image translation. The authentication center generates encryption and decryption keys by training a cyclic image translation model. The data owner and data user obtain data upload and download permissions on the cloud server. The data owner submits an identity token to the authentication center and completes identity authentication. The data owner uses the encryption key issued by the authentication center to encrypt the original image and upload the encrypted image to the cloud server. The data user submits the identity token to the authentication center and completes identity authentication. The data user downloads the encrypted image from the cloud server and decrypts the downloaded encrypted image using the decryption key issued by the authentication center to generate a high-quality decrypted image. The above method uses deep learning technology to solve the access control problem of cloud data, ensuring that the encrypted image content is imperceptible and the decrypted image has high reconstruction quality.

Description

Technical Field

[0001] The present invention relates to the field of deep learning and cloud data security technology, and in particular to a cloud data access control method based on image translation. Background Art

[0002] Due to limited resources, data owners sometimes upload relatively large amounts of data to cloud servers for storage to manage and maintain outsourced data. However, under the cloud storage model, data owners do not have sufficient management rights over outsourced data, which can easily lead to data privacy leaks.

[0003] Therefore, protecting cloud data security and privacy through effective access control methods is an issue that needs to be addressed urgently. Summary of the Invention

[0004] The purpose of this invention is to propose a cloud data access control method based on image translation, which uses deep learning technology to solve the access control problem of cloud data, ensure the imperceptibility of encrypted image content and the high reconstruction quality of decrypted image, thereby enhancing the privacy protection performance of cloud data.

[0005] The purpose of the present invention is achieved through the following technical solutions:

[0006] A cloud data access control method based on image translation, the method comprising:

[0007] Step 1: The authentication center generates encryption and decryption keys by training a cyclic image translation model.

[0008] Step 2: The data owner and data user complete registration on the cloud server and obtain data upload and download permissions respectively;

[0009] Step 3: The data owner submits the identity token to the authentication center and completes the identity authentication. The authentication center issues the encryption key to the data owner through a secure channel.

[0010] Step 4: The data owner encrypts the original image using the encryption key issued by the authentication center and uploads the encrypted image to the cloud server;

[0011] Step 5: The data user submits the identity token to the authentication center and completes the identity authentication. The authentication center issues the decryption key to the data user through a secure channel.

[0012] In step 6, the data user downloads the encrypted image from the cloud server and decrypts the downloaded encrypted image using the decryption key issued by the authentication center to generate a high-quality decrypted image.

[0013] It can be seen from the technical solution provided by the above-mentioned present invention that the above-mentioned method uses deep learning technology to solve the access control problem of cloud data, ensures the imperceptibility of the encrypted image content and the high reconstruction quality of the decrypted image, thereby enhancing the privacy protection performance of cloud data. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0015] Figure 1 A schematic diagram of the flow of a cloud data access control method based on image translation provided by an embodiment of the present invention;

[0016] Figure 2 Schematic diagram of the structure of the generative network and the discriminative network in the cyclic image translation model according to an embodiment of the present invention;

[0017] Figure 3 Four original images and their corresponding category labels are randomly selected from the CIFAR-10 dataset for the examples given in this invention;

[0018] Figure 4 The encryption key used in the examples of the present invention is The encrypted image generated by encrypting the original image and the corresponding PSNR value;

[0019] Figure 5 The encryption key used for the examples given in this invention is The encrypted image generated by encrypting the original image and the corresponding PSNR value;

[0020] Figure 6 The decryption key used in the examples of this invention is Encryption key-based The decrypted image and the corresponding PSNR value generated by decrypting the encrypted image;

[0021] Figure 7 The decryption key used in the examples of this invention is Encryption key-based The decrypted image and the corresponding PSNR value generated by decrypting the encrypted image;

[0022] Figure 8 The decryption key used in the examples of this invention is Encryption key-based The decrypted image and the corresponding PSNR value generated by decrypting the encrypted image;

[0023] Figure 9 The decryption key used in the examples of this invention is Encryption key-based The encrypted image is decrypted to generate the decrypted image and the corresponding PSNR value. DETAILED DESCRIPTION

[0024] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments, and do not constitute a limitation of the present invention. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0025] like Figure 1 FIG2 is a flow chart of a cloud data access control method based on image translation provided by an embodiment of the present invention, wherein the method includes:

[0026] Step 1: The authentication center generates encryption and decryption keys by training a cyclic image translation model.

[0027] In this step, the specific process is:

[0028] (1) The authentication center has the same original image set x as the data owner, and performs block segmentation and random block scrambling operations on all images in the original image set x to obtain the target image set Where k is the index of the image block size and random block scrambling order, and there are K different image block sizes and random block scrambling orders;

[0029] (2) The authentication center checks the original image and the target image set Construct a cyclic image translation model and initialize the model parameters. The cyclic image translation model consists of two generative networks G with the same structure. A,k and G B,k And two discriminant networks D with the same structure A,k and D B,k constitute;

[0030] Among them, G A,k represents the generation network from the original image domain to the target image domain; G B,k represents the generation network from the target image domain to the original image domain; D A,k Represents the discriminant network that determines whether the image belongs to the original image domain; D B,k Represents the discriminant network that determines whether the image belongs to the target image domain;

[0031] like Figure 2The figure shows a schematic diagram of the structure of the generative network and the discriminative network in the cyclic image translation model described in an embodiment of the present invention. In the generative network and the discriminative network: the input image is a color image with a size of 32×32; Conv(in_channels, out_channels, kernel_size, stride, padding) is a two-dimensional convolution operation with the number of input channels being in_channels, the number of output channels being out_channels, the convolution kernel size being kernel_size, the convolution kernel step being stride, and the number of padding being padding; ReflectionPad2d(3) is a reflection padding operation with a padding number of 3; ResidualBlock is a residual network module; Upsample(scale_factor=2.0) is an upsampling operation with a scale factor of 2.0; BN is a batch normalization operation; ReLU, Tanh, and LeakyReLU are activation functions in deep learning;

[0032] (3) The authentication center defines the generation adversarial loss function as:

[0033]

[0034] Where x represents the original image in the original image set x; Represents the target image set The target image in G A,k (x) indicates the use of the generative network G A,k The image generated by transforming the original image x; D B,k (G A,k (x)) represents the discriminant image G A,k (x) the probability of belonging to the target image domain; Represents the target image for discrimination The probability of belonging to the target image domain; Indicates the use of the generative network G B,k For the target image The image generated by the transformation; Represents the discriminant image The probability of belonging to the original image domain; D A,k (x) represents the probability of judging that the original image x belongs to the original image domain; Represents the expectation operation based on the statistical distribution of the original image; Represents the expectation operation based on the statistical distribution of the target image;

[0035] (4) The authentication center defines the cycle consistency loss function Expressed as:

[0036]

[0037] Among them, G B,k (G A,k (x)) indicates the use of the generative network G B,k For image G A,k (x) transform the generated image; Indicates the use of the generative network G A,k For images The image generated by the transformation;

[0038] (5) The certification center defines the loss function of the cyclic image translation model Expressed as:

[0039]

[0040] (6) The authentication center pairs K original images and target images. Independently train the cyclic image translation model, that is, minimize the loss function of the cyclic image translation model To obtain the corresponding generative network and discriminant network Expressed as:

[0041]

[0042] The resulting generative network As the encryption key, the resulting generated network as a decryption key; Indicates that the optimal generation network is obtained by adversarial training of the generation network and the discrimination network and Operation.

[0043] Step 2: The data owner and data user complete registration on the cloud server and obtain data upload and download permissions respectively;

[0044] In this step, the data owner and the data user submit the registration credentials Do to the cloud server respectively. k ,1≤k≤K and User k ,1≤k≤K, respectively obtain the upload permission and download permission of the data.

[0045] Step 3: The data owner submits the identity token to the authentication center and completes the identity authentication. The authentication center issues the encryption key to the data owner through a secure channel.

[0046] In this step, you have the registration credentials Do k ,1≤k≤K The data owner submits the identity token Do_Token to the authentication center k , the authentication center verifies the identity token Do_Tokenk effectiveness;

[0047] The authentication center issues encryption keys to data owners who have passed identity authentication through a secure channel.

[0048] Step 4: The data owner encrypts the original image using the encryption key issued by the authentication center and uploads the encrypted image to the cloud server;

[0049] In this step, you have the registration credentials Do k ,1≤k≤K The data owner uses the encryption key Encrypt the original image x to generate an encrypted image

[0050] Do have registration credentials k The data owner will encrypt the image Upload to the cloud server.

[0051] Step 5: The data user submits the identity token to the authentication center and completes the identity authentication. The authentication center issues the decryption key to the data user through a secure channel.

[0052] In this step, you have the registered credentials User k ,1≤k≤K The data user submits the identity token User_Token to the authentication center k , the authentication center verifies the identity token User_Token k effectiveness;

[0053] The authentication center issues decryption keys to data users who have passed identity authentication through a secure channel.

[0054] In step 6, the data user downloads the encrypted image from the cloud server and decrypts the downloaded encrypted image using the decryption key issued by the authentication center to generate a high-quality decrypted image.

[0055] In this step, you have the registered credentials User k ,1≤k≤K data users download the encrypted image uploaded by the data owner from the cloud server

[0056] User with registration credentials k The data user uses the decryption key Download encrypted images Decrypt and generate high-quality decrypted images

[0057] It should be noted that the contents not described in detail in the embodiments of the present invention belong to the prior art known to those skilled in the art.

[0058] In order to test the privacy protection performance of the method described in the embodiment of the present invention, the following two sets of encryption / decryption keys are used. and The encryption / decryption image quality in the following case is used as an example to illustrate:

[0059] The privacy protection performance was tested using the CIFAR-10 dataset. Adam was used as the optimization algorithm for the generative and discriminative networks in the cyclic image translation model (with parameter beta1 set to 0.5 and parameter beta2 set to 0.999). The cyclic image translation model was trained with 50 epochs, a batch size of 64, and a learning rate of 0.0002. Peak Signal-to-Noise Ratio (PSNR) was used to measure the quality of encrypted and decrypted images, measured in dB.

[0060] like Figure 3 4 original images randomly selected from the CIFAR-10 dataset and their corresponding category labels are shown in the example of the present invention, which are images of airplanes, cars, cats, and horses.

[0061] like Figure 4 The example of the present invention uses the encryption key The encrypted image generated by encrypting the original image and the corresponding PSNR value (PSNR is Peak Signal-to-Noise Ratio, translated as Peak Signal-to-Noise Ratio), such as Figure 5 The example of the present invention uses the encryption key The encrypted image and the corresponding PSNR value generated by encrypting the original image. Figure 4 and Figure 5 It can be seen that after encryption, the content of the encrypted image can no longer be perceived.

[0062] like Figure 6 The example of the present invention uses the decryption key Encryption key-based The decrypted image generated by decrypting the encrypted image and the corresponding PSNR value are as follows: Figure 7 The example of the present invention uses the decryption key Encryption key-based The decrypted image and the corresponding PSNR value are generated by decrypting the encrypted image. Figure 6 and Figure 7It can be seen that the decrypted image generated using the correct decryption key has a high reconstruction quality and can perceive the content of the decrypted image.

[0063] like Figure 8 The example of the present invention uses the decryption key Encryption key-based The decrypted image generated by decrypting the encrypted image and the corresponding PSNR value are as follows: Figure 9 The example of the present invention uses the decryption key Encryption key-based The decrypted image and the corresponding PSNR value are generated by decrypting the encrypted image. Figure 8 and Figure 9 It can be seen that the decrypted image generated using the wrong decryption key has a lower reconstruction quality and cannot be perceived.

[0064] In summary, the method described in the embodiment of the present invention trains a cyclic image translation model to generate encryption keys and decryption keys, and performs encryption and decryption operations on the data owner's original image based on key distribution, while ensuring that the encrypted image content is imperceptible and the decrypted image has high reconstruction quality, thereby enhancing the privacy protection performance of cloud data.

[0065] In addition, those skilled in the art will understand that all or part of the steps in the above-mentioned embodiment method can be implemented by instructing the relevant hardware through a program, and the corresponding program can be stored in a computer-readable storage medium. The above-mentioned storage medium can be a read-only memory, a disk or an optical disk, etc.

[0066] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims. The information disclosed in the background technology section of this article is only intended to deepen the understanding of the overall background technology of the present invention, and should not be regarded as an admission or any form of implication that the information constitutes prior art already known to those skilled in the art.

Claims

1. A cloud data access control method based on image translation, characterized in that: The method comprises: Step 1: The authentication center generates encryption and decryption keys by training a cyclic image translation model. The process of step 1 is specifically as follows: (1) The authentication center has the same original image set x as the data owner, and performs block segmentation and random block scrambling operations on all images in the original image set x to obtain the target image set 1≤k≤K; where k is the index of the image block size and random block scrambling order, and there are K different image block sizes and random block scrambling orders; (2) The authentication center checks the original image and the target image set 1≤k≤K constructs a cyclic image translation model and initializes the model parameters. The cyclic image translation model consists of two generative networks G with the same structure. A,k and G B,k , and two discriminant networks D with the same structure A,k and D B,k constitute; Among them, G A,k represents the generation network from the original image domain to the target image domain; G B,k represents the generation network from the target image domain to the original image domain; D A,k Represents the discriminant network that determines whether the image belongs to the original image domain; D B,k Represents the discriminant network that determines whether the image belongs to the target image domain; (3) The authentication center defines the generation adversarial loss function as: Where x represents the original image in the original image set x; Represents the target image set The target image in G A,k (x) indicates the use of the generative network G A,k The image generated by transforming the original image x; D B,k (G A,k (x)) represents the discriminant image G A,k (x) the probability of belonging to the target image domain; Represents the target image for discrimination The probability of belonging to the target image domain; Indicates the use of the generative network G B,k For the target image The image generated by the transformation; Represents the discriminant image The probability of belonging to the original image domain; D A,k (x) represents the probability of judging that the original image x belongs to the original image domain; Represents the expectation operation based on the statistical distribution of the original image; Represents the expectation operation based on the statistical distribution of the target image; (4) The authentication center defines the cycle consistency loss function L cyc (G A,k ,G B,k ) is expressed as: Among them, G B,k (G A,k (x)) indicates the use of the generative network G B,k For image G A,k (x) transform the generated image; Indicates the use of the generative network G A,k For images The image generated by the transformation; (5) The certification center defines the loss function of the cyclic image translation model Expressed as: (6) The authentication center pairs K original images and target images. 1≤k≤K independently train the cyclic image translation model, that is, minimize the loss function of the cyclic image translation model To obtain the corresponding generative network and discriminant network Expressed as: The resulting generative network As the encryption key, the resulting generated network as a decryption key; Indicates that the optimal generation network is obtained by adversarial training of the generation network and the discrimination network and Operations; Step 2: The data owner and data user complete registration on the cloud server and obtain data upload and download permissions respectively; Step 3: The data owner submits the identity token to the authentication center and completes the identity authentication. The authentication center issues the encryption key to the data owner through a secure channel. Step 4: The data owner encrypts the original image using the encryption key issued by the authentication center and uploads the encrypted image to the cloud server; Step 5: The data user submits the identity token to the authentication center and completes the identity authentication. The authentication center issues the decryption key to the data user through a secure channel. In step 6, the data user downloads the encrypted image from the cloud server and decrypts the downloaded encrypted image using the decryption key issued by the authentication center to generate a high-quality decrypted image.

2. The cloud data access control method based on image translation according to claim 1 is characterized in that: The process of step 3 is specifically as follows: Do have registration credentials k ,1≤k≤K The data owner submits the identity token Do_Token to the authentication center k , the authentication center verifies the identity token Do_Token k effectiveness; The authentication center issues encryption keys to data owners who have passed identity authentication through a secure channel.

3. The cloud data access control method based on image translation according to claim 2 is characterized in that: The process of step 4 is specifically as follows: Do have registration credentials k ,1≤k≤K The data owner uses the encryption key Encrypt the original image x to generate an encrypted image Do have registration credentials k The data owner will encrypt the image Upload to the cloud server.

4. The cloud data access control method based on image translation according to claim 1 is characterized in that: The process of step 5 is specifically as follows: User with registration credentials k ,1≤k≤K The data user submits the identity token User_Token to the authentication center k , the authentication center verifies the identity token User_Token k effectiveness; The authentication center issues decryption keys to data users who have passed identity authentication through a secure channel.

5. The cloud data access control method based on image translation according to claim 4 is characterized in that: The process of step 6 is specifically as follows: User with registration credentials k ,1≤k≤K data users download the encrypted image uploaded by the data owner from the cloud server User with registration credentials k The data user uses the decryption key Download encrypted images Decrypt and generate high-quality decrypted images

Images