Safe outsourcing image retrieval method and device, electronic equipment and medium

Abstract

The embodiment of the invention relates to the technical field of image retrieval, and provides a secure outsourcing image retrieval method and device, electronic equipment and a medium, and the method comprises the following steps: obtaining an image retrieval request sent by a query user, including retrieving an encrypted image; in response to the image retrieval request, performing ciphertext feature extraction on the retrieved encrypted image to obtain a first orthogonal moment feature vector, and calculating a first ciphertext hash value corresponding to the first orthogonal moment feature vector; determining a target hash bucket in a hash bucket set based on the first ciphertext hash value; screening a preset number of target encrypted images with the minimum distance from the first orthogonal moment feature vector in the target hash bucket; and converting the encryption form of the target encrypted image, and returning the converted target encrypted image to the query user. Therefore, efficient retrieval of the secure outsourcing image is realized, and the security of the scheme is improved.

Description

Technical Field

[0001] This invention relates to the field of image retrieval technology, and in particular to a secure outsourced image retrieval method, apparatus, electronic device, and medium. Background Technology

[0002] With the development of cloud technology, more and more image owners, including individuals and businesses, are willing to outsource large amounts of image data to remote, "honest but curious" cloud servers to save on local storage and computing costs. However, due to the open and semi-trusted nature of cloud servers, malicious entities, including the cloud servers themselves, may steal sensitive information contained in the images, such as personal identification, home address, and workplace.

[0003] To protect image privacy and facilitate image sharing, secure outsourced image retrieval has been extensively researched. However, existing secure outsourced image retrieval systems use shared keys and do not support key revocation, making it impossible to dynamically delete image owners and query users. This results in weak scalability for dynamically updating image owners and query users, failing to meet the needs of multi-source, multi-user scenarios. Moreover, in the event of unauthorized operations, the inability to revoke the key held by an unauthorized user will cause significant losses to legitimate users. Furthermore, the index of secure outsourced image retrieval systems is built by the image owner. Once the image owner needs to upload or revoke image data in the cloud, they must incur significant communication and computational overhead to update the index, further weakening the scalability of dynamically updating image data. In addition, after query users decrypt the retrieved encrypted images to obtain plaintext images, they may illegally redistribute the plaintext images, leading to the risk of image infringement.

[0004] Existing solutions employ encrypted domain information hiding technology, which compromises the content of the original image. Therefore, how to protect image copyright and track unauthorized users while ensuring image data integrity has become a pressing issue. Summary of the Invention

[0005] This invention provides a secure outsourced image retrieval method, apparatus, electronic device, and medium to address the shortcomings of existing technologies, such as the damage to the original image content caused by the use of encrypted domain information hiding technology and the low security of key sharing. It achieves efficient retrieval of secure outsourced images, copyright protection, and tracking of unauthorized users, thereby improving the security of the solution.

[0006] This invention provides a secure outsourced image retrieval method, comprising: Obtain an image retrieval request sent by a querying user, wherein the image retrieval request carries a retrieval encrypted image generated based on the query image; In response to the image retrieval request, the encrypted image to be retrieved is subjected to ciphertext feature extraction to obtain the first orthogonal moment feature vector corresponding to the encrypted image to be retrieved, and the first ciphertext hash value corresponding to the first orthogonal moment feature vector is calculated; Based on the first ciphertext hash value, a target hash bucket with the same first ciphertext hash value is determined in the hash bucket set, wherein the hash bucket set is an image index pre-constructed by encrypting the original image set; Filter a preset number of target encrypted images within the target hash bucket that have the smallest distance to the first orthogonal moment feature vector; The encryption format of the target encrypted image is converted, and the converted target encrypted image is returned to the querying user.

[0007] In one possible implementation, the method further includes: Obtain the encrypted image set obtained by the image owner encrypting the original image using the owner's key; The ciphertext domain features of the encrypted image set are extracted using a fractional orthogonal moment feature extraction method to obtain a second orthogonal moment feature vector, and the second ciphertext hash value corresponding to the second orthogonal moment feature vector is calculated. Encrypted images with the same hash value as the second ciphertext are mapped to the same hash bucket to obtain a set of hash buckets; An image index is constructed based on the set of hash buckets.

[0008] In one possible implementation, the method further includes: The first cloud server and the second cloud server collaboratively extract the ciphertext domain features of the retrieved encrypted image through a secure multi-party computation protocol. The first cloud server holds the system public key and a first part of the decryption key, while the second cloud server holds the system public key and a second part of the decryption key. Neither cloud server can obtain the plaintext data independently. The secure multi-party computation protocol includes a fast cross-domain secure multiplication protocol and a fast secure addition protocol. The fast cross-domain secure multiplication protocol is used to implement the multiplication operation of two ciphertext data, and the fast secure addition protocol is used to implement the addition operation of two ciphertext data.

[0009] In one possible implementation, the method further includes: The first cloud server selects four random integers, encrypts two of the random integers using the query user key, and encrypts the other two random integers using the system public key; The first cloud server performs a homomorphic operation on the ciphertext data to be multiplied of the retrieved encrypted image and an encrypted random integer to obtain first blinded ciphertext data. After performing a first partial decryption on the first blinded ciphertext data using the first part of the decryption key, the first partially decrypted first blinded ciphertext data and the encrypted random integer are sent to the second cloud server. The second cloud server performs a second decryption using the second part of the decryption key to obtain first plaintext data with a random number mask. The first plaintext data is then re-encrypted using the system public key to obtain the first plaintext data ciphertext. The first cloud server performs a multiplication operation on the first plaintext data ciphertext to remove the influence of random numbers and obtain the ciphertext multiplication result.

[0010] In one possible implementation, the method further includes: The first cloud server selects two random integers and encrypts them using the system public key. The first cloud server performs a homomorphic operation on the data to be encrypted of the retrieved encrypted image and the encrypted random integer to obtain the second blinded ciphertext data; The first cloud server uses the first part of the decryption key to perform a first partial decryption of the second blinded ciphertext data and then sends it to the second cloud server. The second cloud server uses the second part of the decryption key to perform a second decryption, and adds the results of the second decryption to obtain the second plaintext data with a random number mask. The second plaintext data is then re-encrypted using the system public key to obtain the second plaintext data ciphertext. The first cloud server performs a multiplication operation on the second plaintext data ciphertext to remove the influence of random numbers, and obtains the ciphertext addition result.

[0011] In one possible implementation, the method further includes: Based on the results of ciphertext multiplication and ciphertext addition, the ciphertext domain features of the retrieved encrypted image are extracted, and the ciphertext domain features are converted into feature vectors to obtain the first orthogonal moment feature vector.

[0012] In one possible implementation, the method further includes: After returning the converted target encrypted image to the querying user, the target encrypted image is watermarked by the watermark authorization center to obtain a copyright verification certificate. The copyright verification certificate is used to determine whether there is any image infringement.

[0013] The present invention also provides a secure outsourced image retrieval device, comprising the following modules: The acquisition module is used to acquire image retrieval requests sent by querying users, wherein the image retrieval requests carry a retrieval encrypted image generated based on the query image; The feature extraction module is used to extract ciphertext features from the retrieved encrypted image in response to the image retrieval request, obtain the first orthogonal moment feature vector corresponding to the retrieved encrypted image, and calculate the first ciphertext hash value corresponding to the first orthogonal moment feature vector; The retrieval module is used to determine the target hash bucket that is the same as the first ciphertext hash value in the hash bucket set based on the first ciphertext hash value, wherein the hash bucket set is an image index that is pre-constructed by encrypting the original image set; The retrieval module is used to filter a preset number of target encrypted images within the target hash bucket that have the smallest distance to the first orthogonal moment feature vector; The retrieval module is used to convert the encryption form of the target encrypted image and return the converted target encrypted image to the querying user.

[0014] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the secure outsourcing image retrieval method as described above.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the secure outsourcing image retrieval method as described above.

[0016] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the secure outsourcing image retrieval method as described above.

[0017] This invention provides a secure outsourced image retrieval method, apparatus, electronic device, and medium. The method involves: acquiring an image retrieval request sent by a query user, wherein the image retrieval request carries a retrieval encrypted image generated based on the query image; extracting ciphertext features from the retrieval encrypted image in response to the image retrieval request to obtain a first orthogonal moment feature vector corresponding to the retrieval encrypted image, and calculating a first ciphertext hash value corresponding to the first orthogonal moment feature vector; determining target hash buckets with the same first ciphertext hash value from a hash bucket set, wherein the hash bucket set is an image index pre-constructed after encryption based on the original image set; selecting a predetermined number of target encrypted images within the target hash buckets that have the smallest distance to the first orthogonal moment feature vector; converting the encryption form of the target encrypted images; and returning the converted target encrypted images to the query user. Compared to the shortcomings of existing technologies that use encrypted domain information hiding technology, which damages the content of the original image and suffers from low key sharing security, this solution achieves efficient retrieval of secure outsourced images and improves the security of the solution. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the structure of the secure outsourced image retrieval system provided by the present invention.

[0020] Figure 2 This is a flowchart illustrating the secure outsourced image retrieval method provided by the present invention.

[0021] Figure 3 This is a schematic diagram of the structure of the secure outsourced image retrieval device provided by the present invention.

[0022] Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0024] To facilitate understanding of the embodiments of the present invention, further explanations and descriptions will be provided below with reference to the accompanying drawings and specific embodiments. These embodiments do not constitute a limitation on the embodiments of the present invention.

[0025] Figure 1 This is a schematic diagram of the secure outsourced image retrieval system provided by the present invention. This system is a secure outsourced image retrieval system in a cloud computing environment, adopting a dual-server collaborative architecture, and mainly includes six core components: The Trusted Key Generation Center (KGC), acting as the system's root of trust, employs a distributed dual-threshold public-key cryptography method to generate independent key pairs for each participant. The KGC generates public-private key pairs for each image owner, each querying user, the system public key and partial decryption keys for the first cloud server, the second cloud server, and a watermark management key for the watermark authorization center. Strict isolation of keys among participants eliminates the risk of key sharing at its source.

[0026] The image owner's terminal is responsible for image preprocessing and uploading. The image owner uses their owner's key to encrypt the original image set, generating an encrypted image set, which is then uploaded to the first cloud server. This encryption operation is performed locally, ensuring that the original images are in an encrypted state before transmission.

[0027] The first and second cloud servers form a dual-server secure computing cluster. The first cloud server is the primary computing node, holding the system public key and part of the decryption key; the second cloud server is a collaborative computing node, holding the system public key and another part of the decryption key. The two servers collaborate to complete the ciphertext domain computation task through a fast cross-domain secure multiplication protocol and a fast secure addition protocol. In the image outsourcing stage, the two servers perform fractional-order orthogonal Pollazzek-Fourier moment feature extraction on the encrypted image. After scaling and quantizing the coefficients, radial kernel function, and circular kernel function as a whole, they use a secure multi-party computation protocol to extract the ciphertext features of the encrypted image set, calculate the ciphertext hash value, and construct a dynamically updatable hash bucket index. In the retrieval stage, the two servers collaboratively extract the ciphertext features of the retrieved encrypted image, calculate the ciphertext hash value, calculate the secure squared Euclidean distance, and perform key conversion.

[0028] The querying user terminal acts as the retrieval initiator. The querying user uses their public key to encrypt the queried image, generating a retrieval trapdoor, and uploads it to the first cloud server. After the retrieval is complete, the querying user uses their private key to decrypt the returned encrypted result, obtaining the plaintext image.

[0029] The Watermarking Authorization Center is responsible for copyright protection and traceability. The center receives the image ID, owner ID, query user ID, and image feature vector encrypted with the center's public key from the search results. Through feature decryption, binarization, chaotic sequence modulation, and generalized Arnold transform, a zero-watermark image is generated without embedding the original image. This is then combined with a digital timestamp and a one-way hash function to construct an immutable copyright verification credential. When a suspected infringing image is discovered, the center uses the same process to generate a verification watermark and compares it with the stored credential, achieving lossless copyright verification and tracing of illegal distributors. The key conversion mechanism is the core of the system's secure multi-user sharing. Through dual-server collaboration, decryption is followed by immediate re-encryption, converting the image from owner-key encrypted form to user-key encrypted form. This ensures that neither party can obtain the plaintext during the conversion process, achieving secure distribution of search results under key isolation. All modules of the system communicate through public channels, maintaining encrypted states throughout all data flow. This protects the privacy of image content and query intent while achieving efficient similarity retrieval and traceable copyright protection.

[0030] Figure 2 This is a flowchart illustrating the secure outsourced image retrieval method provided by the present invention, as shown below. Figure 2 As shown, the method includes the following: First, key generation is performed: Trusted key generation center The distributed two-threshold public-key cryptography method is used to provide security for the image owner, authorized legitimate query users, and the first cloud server within the system. Second cloud server The watermark authorization center generates the corresponding key. , ), ( , and the system public key Partial decryption key and 、( , , , , ).

[0031] Furthermore, Will , Send to the image owner ; Will , Send to authorized and legitimate query users ; System public key Partial decryption key Send to the first cloud server , system public key Partial decryption key Send to the second cloud server ; Send the image owner's public key and query user's public key Give First Cloud Server Second cloud server ; Will , , , , Send it to a trusted watermark authorization center.

[0032] Furthermore, the encrypted image set obtained by the image owner encrypting the original image with the owner's key is obtained; the ciphertext domain features of the encrypted image set are extracted using the fractional orthogonal moment feature extraction method to obtain the second orthogonal moment feature vector, and the second ciphertext hash value corresponding to the second orthogonal moment feature vector is calculated; encrypted images with the same second ciphertext hash value are mapped to the same hash bucket to obtain a hash bucket set; and an image index is constructed based on the hash bucket set.

[0033] Specifically, image owner Use the key to access the image set you have Encryption to obtain encrypted image set and encrypt the image set Send to the remote first cloud server The first cloud server performs encrypted feature extraction on the encrypted image set.

[0034] The specific encryption feature extraction process is as follows: First, to enable features to have strong image representation capabilities, a new feature extraction method called fractional orthogonal Pollazzek-Fourier moment features is proposed.

[0035] Using Pollaczek polynomials ,in, , , , and These are free parameters. It satisfies orthogonality in the interval from -1 to 1 in the Cartesian coordinate system. A new polynomial obtained by shifting and weighting it has a large number of uniformly distributed zeros and is called the weighted radially shifted orthogonal Pollazzek polynomial. , It is orthogonal on the unit disk in the polar coordinate system between 0 and 1.

[0036] Furthermore, to improve the performance of the constructed features in image representation and enhance their noise resistance in image retrieval, a weighted radial translation orthogonal Pollazzek polynomial is used. The concept of fractional order is introduced, and fractional order weighted radial translation orthogonal Pollazzek polynomials are constructed. This is then used as the radial kernel function of the new image orthogonal moment feature. To ensure that the constructed new fractional moment feature is inherently rotation-invariant and flip-invariant, a new type of feature called the fractional orthogonal Pollazzek-Fourier moment feature is constructed. The details are as follows: The first cloud server uses data scaling technology to measure the coefficients of the orthogonal moment features of the image. Radial kernel function and circular kernel function As a whole After calculation, multiplying by a large integer completes the scaling process to ensure the effectiveness of feature extraction. Then, data quantization techniques are used to quantize and round the scaled result to obtain the final value. This ensures that the quantized data meets the integer data requirements of distributed dual-threshold public-key cryptography methods.

[0037] Then, the first cloud server uses the Fast Cross-Domain Secure Multiplication Protocol and the Fast Secure Addition Protocol to achieve secure and efficient extraction of orthogonal moment features from the image.

[0038] Specifically, firstly, the first cloud server selects four random integers. , , and Using the image owner public key encryption and Get encrypted random integer and Using the system public key encryption and Get encrypted random integer and Next, the first cloud server performs calculations. , , as well as and use the first part of the decryption key After decryption, the decryption result will be... , , , Together , , , The encrypted data is sent to the second cloud server. Upon receiving the encrypted data, the second cloud server uses the second part of the decryption key. Decrypt it to obtain , and Next, the second cloud server uses the system public key. To each , and After encryption, the ciphertext , and The encrypted result was sent to the first cloud server, and the first cloud server received the encrypted result. , and Then, calculate Remove random integers , , and To obtain the final .

[0039] Finally, to avoid multiple multiplications of large encrypted data, a fast and secure addition protocol is designed to extract orthogonal moment features from images. First, the first cloud server selects two random integers... and Use system public key Encryption obtained and Then, the first cloud server calculates... and Use the first part of the decryption key After decryption, the decryption result is sent to the second cloud server. Upon receiving the encrypted data, the second cloud server uses the second part of the decryption key. Decrypt it, add the decrypted results together, and then use the system public key. The encryption is performed, and the ciphertext is sent to the first cloud server. After receiving the ciphertext result, the first cloud server removes the random integer. and get Repeating the above process, the first cloud server eventually obtains the public key. Orthogonal moment features of encrypted ciphertext images.

[0040] Further, build the index: The first and second cloud servers utilize the extracted orthogonal moment feature vectors of the encrypted image. Calculate its corresponding ciphertext hash value ,in It is an orthogonal moment characteristic The converted ciphertext orthogonal moment eigenvectors Represents the image feature vector value. It is a random vector. From the interval A value of a variable is randomly selected from within. It is a positive integer that is set by the user.

[0041] The first and second cloud servers use a pre-designed secure comparison protocol to analyze the orthogonal moment feature vector of the encrypted image. Encrypted images with corresponding equal hash values ​​are mapped to the same hash bucket to securely build efficient indexes.

[0042] When an image owner uploads a new image, the cloud server performs the above steps and updates the index by placing the hash value of the new image into the corresponding hash bucket. When an existing image needs to be deleted, the cloud server deletes the stored image, its corresponding encrypted image orthogonal moment feature, and its corresponding hash value. Therefore, this scheme is highly scalable. When performing image addition or deletion operations, it is not necessary to rebuild the entire index, thus greatly reducing the computational overhead of index construction.

[0043] S21. Obtain the image retrieval request sent by the query user.

[0044] The image retrieval request carries a retrieval encrypted image generated based on the query image.

[0045] Query User To perform a similarity search on encrypted images, you can use your public key. The query image is encrypted, and the generated encrypted image is uploaded to the first cloud server as a retrieval trap.

[0046] S22. In response to the image retrieval request, perform ciphertext feature extraction on the retrieved encrypted image to obtain the first orthogonal moment feature vector corresponding to the retrieved encrypted image, and calculate the first ciphertext hash value corresponding to the first orthogonal moment feature vector.

[0047] The first cloud server, in cooperation with the second server, securely extracts the orthogonal moment feature vector of the encrypted image. The method for extracting orthogonal moment feature vectors from encrypted images described above is similar in steps to the method for extracting orthogonal moment features from encrypted images.

[0048] Specifically, the first cloud server and the second cloud server collaboratively extract the ciphertext domain features of the retrieved encrypted image through a secure multi-party computation protocol.

[0049] Specifically, the first cloud server selects four random integers, encrypts two of them using the query user key, and encrypts the other two random integers using the system public key. The first cloud server performs a homomorphic operation on the ciphertext data to be multiplied from the retrieved encrypted image and the encrypted random integers to obtain first blinded ciphertext data. After performing a first partial decryption on the first blinded ciphertext data using a first part of the decryption key, the first partially decrypted first blinded ciphertext data and the encrypted random integers are sent to the second cloud server. The second cloud server performs a second decryption using a second part of the decryption key to obtain first plaintext data with a random number mask. The first plaintext data is then re-encrypted using the system public key to obtain the first plaintext data ciphertext. The first cloud server performs a multiplication operation on the first plaintext data ciphertext to remove the influence of the random number, obtaining the ciphertext multiplication result.

[0050] The first cloud server selects two random integers and encrypts them using the system public key. The first cloud server then performs a homomorphic operation on the data to be encrypted from the retrieved encrypted image and the encrypted random integers to obtain second blinded ciphertext data. The first cloud server uses a first part of the decryption key to perform a partial decryption of the second blinded ciphertext data and sends it to the second cloud server. The second cloud server uses a second part of the decryption key to perform a second decryption, and adds the results of the two decryptions to obtain second plaintext data with a random number mask. The first cloud server then re-encrypts the second plaintext data using the system public key to obtain second plaintext ciphertext. Finally, the first cloud server performs a multiplication operation on the second plaintext ciphertext to remove the influence of the random number, obtaining the ciphertext addition result.

[0051] Finally, based on the ciphertext multiplication and addition results, the ciphertext domain features of the retrieved encrypted image are extracted, and the ciphertext domain features are converted into feature vectors to obtain the second orthogonal moment feature vector. .

[0052] Furthermore, the homomorphic property is used to calculate... Corresponding ciphertext hash value ,in, This indicates a query for image feature vector values. It is a random vector. From the interval A value of a variable is randomly selected from within. It is a positive integer that is set by the user.

[0053] S23. Based on the first ciphertext hash value, determine the target hash bucket in the hash bucket set that is the same as the first ciphertext hash value.

[0054] The hash bucket set is an image index that is pre-built by encrypting the original image set; the index building process has been described above.

[0055] The encrypted hash values ​​are compared using the designed secure comparison protocol, and the result is selected based on the hash value. Same target hash bucket.

[0056] S24. Select a preset number of target encrypted images within the target hash bucket that have the smallest distance to the first orthogonal moment feature vector.

[0057] The designed secure squared Euclidean distance protocol is used to calculate the encrypted image orthogonal moment feature vector of the retrieved image. The secure squared Euclidean distance between the orthogonal moment eigenvectors of the encrypted images within the bucket and the corresponding eigenvectors is calculated, and a secure comparison protocol is used to select the images with the smallest corresponding eigenvector distance. A target encrypted image.

[0058] S25. Convert the encryption form of the target encrypted image and return the converted target encrypted image to the querying user.

[0059] Finally, the first cloud server, in cooperation with the second cloud server, Image owner's public key Encrypted images are converted using a secure key conversion protocol to use the user's public key. Encrypted images, and An encrypted image is returned to the user, who can then decrypt the encrypted image using their private key to obtain the corresponding plaintext image.

[0060] Optionally, after returning the converted target encrypted image to the querying user, the target encrypted image can also be watermarked through the watermark authorization center to obtain a copyright verification certificate; the copyright verification certificate is then used to determine whether there is any image infringement.

[0061] Specifically, the watermark generation process is as follows: The first cloud server will retrieve The orthogonal moment features of each image are converted into a public key using a key conversion protocol. The encrypted feature vector is then sent to the watermark authorization center along with the image ID, image owner ID, and query user ID.

[0062] After receiving the information, the watermarking authorization center first determines whether the image copyright information has been registered based on the image ID. If it has been registered, the query user ID is recorded; otherwise, a zero-watermark image is constructed and the query user ID is recorded for copyright protection against illegal secondary distribution of the image.

[0063] The watermark licensing center generates a size based on the image owner's ID. Binary watermark image Then on The scrambled binary image is obtained by performing a generalized Arnold transform. The key used is .

[0064] Watermarking Authorization Center uses private keys The orthogonal moment feature vector of the image is obtained by decrypting the encrypted image orthogonal moment feature vector sent by the cloud server.

[0065] Randomly select from the feature vectors Each feature is analyzed, and its magnitude is calculated to form a new feature vector. .

[0066] For eigenvectors Binarization yields a new binary feature vector. Quantization threshold adopted The median.

[0067] Generating chaotic sequences using asymmetric Tent mapping The corresponding key is Then, using quantization features Chaotic sequences and vector shift operations to generate a size of Two-dimensional feature image and feature image The scrambled feature image is obtained by scrambling using the generalized Arnold transform. The corresponding key is , With scrambled watermark images Perform an XOR operation to obtain a zero-watermark image. .

[0068] The Watermarking Authorization Center uses a one-way hash function to transfer the key. , , With zero-watermark images Calculate the hash value of all information together And add a digital timestamp to it to obtain ,Will It serves as the sole proof of the image owner's copyright information.

[0069] Furthermore, the watermark verification process is as follows: The Watermarking Authorization Center uses illegally redistributed images to generate new zero-watermark images. The generation process is the same as the original zero-watermark generation process.

[0070] The Watermarking Authorization Center uses a one-way hash function to transfer the key. , , With zero-watermark images Calculate the hash value of all information together .

[0071] Watermark Licensing Center Comparison and If the values ​​are equal, the image copyright belongs to the registered image owner, and the scope of unauthorized users can be determined by the recorded query user ID; otherwise, it does not belong to the owner. Since no information is actually embedded in the image, lossless copyright protection can be achieved against unauthorized secondary distribution of images.

[0072] This invention provides a secure outsourced image retrieval method. The method involves: acquiring an image retrieval request sent by a querying user, wherein the request carries a retrieval encrypted image generated based on the query image; extracting ciphertext features from the retrieval encrypted image in response to the request, obtaining a first orthogonal moment feature vector corresponding to the retrieval encrypted image, and calculating a first ciphertext hash value corresponding to the first orthogonal moment feature vector; determining target hash buckets with the same first ciphertext hash value from a hash bucket set, wherein the hash bucket set is an image index pre-constructed after encryption based on the original image set; selecting a predetermined number of target encrypted images within the target hash buckets that have the smallest distance to the first orthogonal moment feature vector; converting the encryption form of the target encrypted images; and returning the converted target encrypted images to the querying user. Compared to the shortcomings of existing technologies that use encrypted domain information hiding technology, which damages the content of the original image and suffers from low key sharing security, this method achieves efficient retrieval of secure outsourced images, copyright protection, and tracking of unauthorized querying users, thus improving the security of the solution.

[0073] The secure outsourcing image retrieval device provided by the present invention is described below. The secure outsourcing image retrieval device described below can be referred to in correspondence with the secure outsourcing image retrieval method described above.

[0074] Figure 3 This is a schematic diagram of the structure of the secure outsourced image retrieval device provided by the present invention, specifically including: The acquisition module 301 is used to acquire an image retrieval request sent by the querying user, wherein the image retrieval request carries a retrieval encrypted image generated based on the query image. For detailed explanations, please refer to the relevant descriptions in the above method embodiments, which will not be repeated here.

[0075] The feature extraction module 302 is used to extract ciphertext features from the retrieved encrypted image in response to the image retrieval request, obtain a first orthogonal moment feature vector corresponding to the retrieved encrypted image, and calculate a first ciphertext hash value corresponding to the first orthogonal moment feature vector. For detailed explanation, please refer to the relevant descriptions in the above method embodiments, which will not be repeated here.

[0076] The retrieval module 303 is used to determine, based on the first ciphertext hash value, a target hash bucket in the hash bucket set that has the same hash value as the first ciphertext hash value. The hash bucket set is an image index pre-constructed by encrypting the original image set. For detailed explanations, please refer to the relevant descriptions in the above method embodiments; they will not be repeated here.

[0077] The retrieval module 303 is used to filter a preset number of target encrypted images within the target hash bucket that have the smallest distance to the first orthogonal moment feature vector. For detailed explanations, please refer to the relevant descriptions in the above method embodiments; they will not be repeated here.

[0078] The retrieval module 303 is used to convert the encryption format of the target encrypted image and return the converted target encrypted image to the querying user. For detailed explanations, please refer to the relevant descriptions in the above method embodiments; they will not be repeated here.

[0079] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4 As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 430, and a communications bus 440, wherein the processor 410, the communications interface 420, and the memory 430 communicate with each other through the communications bus 440. The processor 410 can invoke logical instructions in the memory 430 to execute a secure outsourced image retrieval method. This method includes: acquiring an image retrieval request sent by a querying user, wherein the image retrieval request carries a retrieval encrypted image generated based on the query image; in response to the image retrieval request, extracting ciphertext features from the retrieval encrypted image to obtain a first orthogonal moment feature vector corresponding to the retrieval encrypted image, and calculating a first ciphertext hash value corresponding to the first orthogonal moment feature vector; determining a target hash bucket with the same first ciphertext hash value in a hash bucket set, wherein the hash bucket set is an image index pre-constructed after encryption based on the original image set; filtering a preset number of target encrypted images within the target hash bucket that have the smallest distance to the first orthogonal moment feature vector; converting the encryption form of the target encrypted images; and returning the converted target encrypted images to the querying user.

[0080] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0081] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the secure outsourced image retrieval method provided by the above methods. The method includes: obtaining an image retrieval request sent by a query user, wherein the image retrieval request carries a retrieval encrypted image generated based on the query image; in response to the image retrieval request, performing ciphertext feature extraction on the retrieval encrypted image to obtain a first orthogonal moment feature vector corresponding to the retrieval encrypted image, and calculating a first ciphertext hash value corresponding to the first orthogonal moment feature vector; determining a target hash bucket with the same first ciphertext hash value in a hash bucket set based on the first ciphertext hash value, wherein the hash bucket set is an image index pre-constructed after encryption based on the original image set; filtering a preset number of target encrypted images in the target hash bucket that have the smallest distance to the first orthogonal moment feature vector; converting the encryption form of the target encrypted images, and returning the converted target encrypted images to the query user.

[0082] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the secure outsourced image retrieval method provided by the above methods. The method includes: acquiring an image retrieval request sent by a querying user, wherein the image retrieval request carries a retrieval encrypted image generated based on the query image; in response to the image retrieval request, extracting ciphertext features from the retrieval encrypted image to obtain a first orthogonal moment feature vector corresponding to the retrieval encrypted image, and calculating a first ciphertext hash value corresponding to the first orthogonal moment feature vector; determining a target hash bucket with the same first ciphertext hash value in a hash bucket set, wherein the hash bucket set is an image index pre-constructed after encryption based on the original image set; filtering a preset number of target encrypted images within the target hash bucket that have the smallest distance to the first orthogonal moment feature vector; converting the encryption form of the target encrypted images, and returning the converted target encrypted images to the querying user.

[0083] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0084] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0085] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A secure outsourced image retrieval method, characterized in that, include: Obtain an image retrieval request sent by a querying user, wherein the image retrieval request carries a retrieval encrypted image generated based on the query image; In response to the image retrieval request, the encrypted image to be retrieved is subjected to ciphertext feature extraction to obtain the first orthogonal moment feature vector corresponding to the encrypted image to be retrieved, and the first ciphertext hash value corresponding to the first orthogonal moment feature vector is calculated; Based on the first ciphertext hash value, a target hash bucket with the same first ciphertext hash value is determined in the hash bucket set, wherein the hash bucket set is an image index pre-constructed by encrypting the original image set; Filter a preset number of target encrypted images within the target hash bucket that have the smallest distance to the first orthogonal moment feature vector; The encryption format of the target encrypted image is converted, and the converted target encrypted image is returned to the querying user.

2. The method according to claim 1, characterized in that, The method further includes: Obtain the encrypted image set obtained by the image owner encrypting the original image using the owner's key; The ciphertext domain features of the encrypted image set are extracted using a fractional orthogonal moment feature extraction method to obtain a second orthogonal moment feature vector, and the second ciphertext hash value corresponding to the second orthogonal moment feature vector is calculated. Encrypted images with the same hash value as the second ciphertext are mapped to the same hash bucket to obtain a set of hash buckets; An image index is constructed based on the set of hash buckets.

3. The method according to claim 1, characterized in that, The step of extracting ciphertext features from the retrieved encrypted image in response to the image retrieval request includes: The first cloud server and the second cloud server collaboratively extract the ciphertext domain features of the retrieved encrypted image through a secure multi-party computation protocol. The first cloud server holds the system public key and a first part of the decryption key, while the second cloud server holds the system public key and a second part of the decryption key. Neither cloud server can obtain the plaintext data independently. The secure multi-party computation protocol includes a fast cross-domain secure multiplication protocol and a fast secure addition protocol. The fast cross-domain secure multiplication protocol is used to implement the multiplication operation of two ciphertext data, and the fast secure addition protocol is used to implement the addition operation of two ciphertext data.

4. The method according to claim 3, characterized in that, The step of extracting the ciphertext domain features of the retrieved encrypted image collaboratively by the first cloud server and the second cloud server through a secure multi-party computation protocol includes: The first cloud server selects four random integers, encrypts two of the random integers using the query user key, and encrypts the other two random integers using the system public key; The first cloud server performs a homomorphic operation on the ciphertext data to be multiplied of the retrieved encrypted image and an encrypted random integer to obtain first blinded ciphertext data. After performing a first partial decryption on the first blinded ciphertext data using the first part of the decryption key, the first partially decrypted first blinded ciphertext data and the encrypted random integer are sent to the second cloud server. The second cloud server performs a second decryption using the second part of the decryption key to obtain first plaintext data with a random number mask. The first plaintext data is then re-encrypted using the system public key to obtain the first plaintext data ciphertext. The first cloud server performs a multiplication operation on the first plaintext data ciphertext to remove the influence of random numbers and obtain the ciphertext multiplication result.

5. The method according to claim 3, characterized in that, The step of extracting the ciphertext domain features of the retrieved encrypted image collaboratively by the first cloud server and the second cloud server through a secure multi-party computation protocol includes: The first cloud server selects two random integers and encrypts them using the system public key. The first cloud server performs a homomorphic operation on the data to be encrypted of the retrieved encrypted image and the encrypted random integer to obtain the second blinded ciphertext data; The first cloud server uses the first part of the decryption key to perform a first partial decryption of the second blinded ciphertext data and then sends it to the second cloud server. The second cloud server uses the second part of the decryption key to perform a second decryption, and adds the results of the second decryption to obtain the second plaintext data with a random number mask. The second plaintext data is then re-encrypted using the system public key to obtain the second plaintext data ciphertext. The first cloud server performs a multiplication operation on the second plaintext data ciphertext and then removes the influence of random numbers to obtain the ciphertext addition result.

6. The method according to any one of claims 3-5, characterized in that, The step of extracting ciphertext features from the retrieved encrypted image in response to the image retrieval request, to obtain the first orthogonal moment feature vector corresponding to the retrieved encrypted image, includes: Based on the results of ciphertext multiplication and ciphertext addition, the ciphertext domain features of the retrieved encrypted image are extracted, and the ciphertext domain features are converted into feature vectors to obtain the first orthogonal moment feature vector.

7. The method according to claim 1, characterized in that, The method further includes: After returning the converted target encrypted image to the querying user, the target encrypted image is watermarked by the watermark authorization center to obtain a copyright verification certificate. The copyright verification certificate is used to determine whether there is any image infringement.

8. A secure outsourced image retrieval device, characterized in that, include: The acquisition module is used to acquire image retrieval requests sent by querying users, wherein the image retrieval requests carry a retrieval encrypted image generated based on the query image; The feature extraction module is used to extract ciphertext features from the encrypted image in response to the image retrieval request, obtain the first orthogonal moment feature vector corresponding to the encrypted image, and calculate the first ciphertext hash value corresponding to the first orthogonal moment feature vector. The retrieval module is used to determine the target hash bucket that is the same as the first ciphertext hash value in the hash bucket set based on the first ciphertext hash value, wherein the hash bucket set is an image index that is pre-constructed by encrypting the original image set; The retrieval module is used to filter a preset number of target encrypted images within the target hash bucket that have the smallest distance to the first orthogonal moment feature vector; The retrieval module is used to convert the encryption form of the target encrypted image and return the converted target encrypted image to the querying user.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the secure outsourcing image retrieval method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the secure outsourcing image retrieval method as described in any one of claims 1 to 7.

Images