An image supervision platform and method based on privacy intersection and blockchain

By combining the improved Bloom filter of Cuckoo Hash and the unintentional transmission protocol on the consortium blockchain, the problem of low privacy data protection and regulatory efficiency in image supervision scenarios in existing technologies is solved, and efficient and accurate image intersection calculation and supervision are achieved.

CN116842557BActive Publication Date: 2026-11-17ZHEJIANG UNIV
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202310366275.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-04-07
Publication Date
2026-11-17
Estimated Expiration
2043-04-07

AI Technical Summary

Technical Problem

Existing technologies cannot effectively protect the privacy data of users and regulatory agencies in image supervision scenarios involving high privacy and confidentiality. Furthermore, existing privacy intersection methods suffer from low accuracy and low efficiency, making it difficult to improve the efficiency and quality of supervision, especially among classified institutions.

Method used

An image monitoring platform is implemented through a consortium blockchain by combining a Bloom filter based on Cuckoo Hash with unintentional transmission. Cuckoo Hash is used to reduce hash collisions and improve the accuracy of privacy intersection calculations, while the unintentional transmission protocol enables secure image data intersection calculations.

Benefits of technology

It enables accurate calculation of image intersections without disclosing user and regulatory data, improving regulatory efficiency and accuracy, while reducing the impact of hash collisions on privacy-preserving intersections and enhancing regulatory quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116842557B_ABST
    Figure CN116842557B_ABST
Patent Text Reader

Abstract

This invention discloses an image monitoring platform and method based on privacy intersection and blockchain. The method involves: a confidentiality agency calculating fingerprints of sensitive images and uploading them to the blockchain; a regulatory agency generating necessary parameters and encrypting them before sending them to the regulated party; the regulated party calculating fingerprints of user images and filling them into a confusion Bloom filter based on Cuckoo Hash; the regulatory party querying the fingerprints of sensitive images from the blockchain and constructing a regular filter accordingly; a 2-out-of-1 unintentional transmission protocol between the regulatory agency and the regulated party, with the regulatory agency constructing a confusion Bloom filter based on the protocol; and the regulatory agency calculating the intersection of the sensitive image set and the user image set based on the filter to determine whether the regulated party possesses illegal images. This invention solves the data silo problem faced by confidentiality agencies and regulatory agencies through blockchain, while simultaneously improving the accuracy of privacy intersection using a confusion Bloom filter based on Cuckoo Hash, thus enhancing the quality of image monitoring.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the fields of blockchain, privacy computing, and other technologies, and in particular to an image supervision platform and method based on privacy intersection and blockchain. Background Technology

[0002] With the development of the internet, more and more people are using social media platforms such as WeChat, Weibo, and Douban to send or share personal photos. Furthermore, with the popularity of non-fungible tokens (NFTs), many users are using NFT trading platforms to tokenize their private images and put them on the blockchain to form digital assets, aiming to protect image copyrights and facilitate digital asset transactions. However, while these platforms provide convenient image uploading, sharing, and trading services to users, they also present some regulatory challenges. Therefore, how to effectively regulate institutions holding user image libraries to prevent the spread of illegal images has become a growing concern for researchers.

[0003] Currently, technologies for detecting similar images are relatively mature, with common methods including deep learning-based image similarity detection, hash-based image similarity detection, and feature extraction-based image similarity detection. These technologies have wide applications, but are limited to scenarios where the image holders do not have specific requirements for image privacy protection. In regulatory scenarios involving high privacy and confidentiality, both user-owned private images stored on the regulated platform and confidential images held by the regulatory department are considered private data. During image detection, neither party should be able to access the other's private images except for those shared by the regulatory body. In the aforementioned scenarios of regulating illegal and confidential images, the regulators hold a confidential image library used to detect illegal images held by users. Due to the special nature of the source image library held by the regulators, these images cannot be disclosed or sent to the regulated party. Conversely, the regulated party holds a user-owned private image library containing a large amount of user privacy information. To prevent the leakage of user information, these images also cannot be disclosed to the regulators. Therefore, the aforementioned image detection technologies alone cannot meet the image regulation requirements in these specific scenarios.

[0004] To protect the privacy of data from both parties involved in the oversight process, privacy intersection techniques have been introduced into privacy data oversight mechanisms due to their ability to achieve privacy intersection without disclosing the data of either party. Numerous studies on privacy intersection have been conducted both domestically and internationally, with widely used methods including third-party-based privacy intersection, public-key encryption-based privacy intersection, unintended transmission-based privacy intersection, and Bloom filter-based privacy intersection. The advantages of third-party-based privacy intersection methods are obvious, such as the method described in patent CN 114884675 A, "Multi-party Privacy Intersection Method, Apparatus, Device, and Medium Based on Bit Transmission." This method introduces a third party to perform intersection calculations on the privacy datasets held by both parties. This method is easy to implement and understand, but its disadvantage is that it exposes the data of both parties involved in the privacy intersection to a third party, potentially leading to data leakage if the third party is untrustworthy. For public-key encryption-based schemes, this method does not rely on a third party, but the use of a public-private key mechanism and the significant computational overhead of public keys increase the time complexity of privacy intersection. For example, in patent CN 114726611 A, entitled "Method, System, and Apparatus for Multi-Party Privacy Set Intersection," a multi-party privacy intersection algorithm based on additive homomorphic encryption is proposed. This effectively solves the problem of supporting multiple initiators and participants in privacy intersection. However, this method requires constructing public-private key pairs and performing multiple rounds of asymmetric encryption on the dataset, resulting in low efficiency in privacy intersection. Compared with public-key encryption-based schemes, privacy intersection methods based on unintentional transmission offer significant improvements in computational complexity. For instance, patent CN 113343305 A, "Method, Apparatus, Device, and Storage Medium for Calculating the Intersection of Privacy Data," describes a method for calculating the intersection of privacy data based on cuckoo hashing and unintentional transmission protocol, which can address the low efficiency problem of privacy intersection to some extent. However, to improve the efficiency of privacy intersection, this method requires exchanging a large amount of intermediate data when executing the unintentional transmission protocol, resulting in excessive communication volume. By leveraging the filtering capabilities of Bloom filters and combining them with unintentional transmission techniques, privacy-focused intersection schemes based on Bloom filters can further improve the efficiency of privacy intersection. However, due to the inherent limitations of Bloom filters, hash collisions can occur during filter construction, and error rates can arise when determining whether an element is in the set. These issues lead to a decrease in the accuracy of privacy intersection. Patent CN 114911815A, a database SQL query method, discloses a privacy intersection algorithm based on a traditional obfuscated Bloom filter and unintentional transmission. This method utilizes obfuscated Bloom filters to accelerate the efficiency of privacy intersection, but collisions may occur at the same position during filter construction, causing older elements to be overwritten, which significantly reduces the accuracy of privacy intersection.

[0005] Furthermore, some classified institutions and enterprises are currently very large, possessing various types of classified images, but these images cannot be securely and effectively shared among regulatory agencies. Therefore, how to effectively integrate classified institutions and regulatory agencies into the regulatory system to improve the efficiency and quality of supervision is also an urgent problem to be solved. Summary of the Invention

[0006] To address the shortcomings of existing technologies, namely the limitations of traditional obfuscated Bloom filters and privacy intersection techniques based on unintentional transmission, this invention proposes an improved Bloom filter based on Cuckoo Hashing, which is combined with unintentional transmission to achieve privacy intersection of image datasets. By incorporating the idea of ​​Cuckoo Hashing into the Bloom filter, it significantly improves the accuracy of privacy intersection compared to traditional obfuscated Bloom filter-based privacy intersection methods, and also achieves higher efficiency in solving privacy intersections compared to public-key encryption-based methods.

[0007] This invention is achieved through the following technical solution:

[0008] In a first aspect, the present invention provides an image supervision platform based on the intersection of consortium blockchain and privacy, which consists of eight parts: consortium blockchain, image owner, social platform operator, image storage system, NFT trading platform, off-chain storage system, confidentiality agency, and regulatory agency.

[0009] The consortium blockchain is a blockchain with strict access control jointly maintained by a confidentiality agency and a regulatory agency. Only the confidentiality agency and the regulatory agency are allowed to read, write, and send transactions on the data in the consortium blockchain.

[0010] The owners of the images are ordinary users of social media platforms or users who have a need to convert their image works into NFTs;

[0011] The social platform operator is used to operate an online social platform and provide users with social services such as sending and sharing personal pictures; it obtains social pictures shared by users from the image storage system and then calculates the fingerprint of the corresponding picture according to the image fingerprint calculation method;

[0012] The image storage system is a cloud storage system or an offline distributed storage system, used to store personal social images uploaded by users;

[0013] The NFT trading platform is used to provide NFT trading services to users, receive image assets uploaded by users that are to be put on the blockchain, and store them in the off-chain storage system; retrieve the images uploaded by users from the off-chain storage system, and then calculate the fingerprint of the corresponding image according to the image fingerprint calculation method;

[0014] The off-chain storage system is a decentralized storage system used to store the user's original image data before the NFT is uploaded to the blockchain; the decentralized storage system uses the InterPlanetary File System (IPFS), which decentralizes and distributes the user's original image data across different storage nodes.

[0015] The confidential organization is an organization that possesses a source image library that is prohibited from being disclosed. This organization calculates the fingerprint of its own confidential images according to the image fingerprint calculation method, and then uploads it to the consortium blockchain, so that the fingerprint of the image is shared with other organizations on the consortium blockchain, but the original image corresponding to the image fingerprint is not disclosed, and the data on the chain is immutable.

[0016] The regulatory bodies are several regulatory bodies used to supervise the legality of personal images uploaded by users to NFT trading platforms and social platform operators; to read image fingerprints uploaded by confidential institutions from the consortium blockchain, where the image fingerprints come from the same confidential institution or from different confidential institutions on the consortium blockchain; and to supervise NFT trading platforms and social platform operators.

[0017] The image owner uploads the image to the social platform operator or NFT trading platform; the social platform operator stores the user-uploaded images in a distributed image storage system, and the NFT trading platform stores user-uploaded images to be uploaded to the blockchain in the off-chain storage system; the confidentiality organization calculates the fingerprints of all images in its confidential image library and uploads these fingerprints to the consortium blockchain; the social platform operator or NFT trading platform, as the regulated party, and the regulatory organization, as the regulator, conduct a round of communication to negotiate relevant parameters for privacy crossover; the regulated party retrieves user-uploaded images in batches from the image storage system and the off-chain storage system, calculates the fingerprints of these images, and constructs a cuckoo hash-based obfuscated Bloom filter; the regulatory organization retrieves the fingerprint dataset of confidential images uploaded by each confidentiality organization from the consortium blockchain and constructs a regular filter based on this fingerprint data; the two parties involved in the regulation execute an unintentional transmission protocol based on their respective Bloom filters, and the regulator determines whether the regulated party possesses illegal user-uploaded images based on the results of the unintentional transmission.

[0018] Furthermore, the image fingerprint calculation method is any one of the perceptual hash algorithm, differential hash algorithm, average hash algorithm, and wavelet hash algorithm; the average hash algorithm is preferred.

[0019] The average hash algorithm is as follows: First, the color image is converted to grayscale. Then, a filter is used to perform noise reduction processing on the image, and a discrete wavelet transform technique is used to transform the image to obtain an image with a frequency below a threshold. Finally, based on the obtained image with a frequency below the threshold, it is reduced to a size of 16×16, and the average pixel grayscale of the image with a frequency below the threshold is calculated. Then, each pixel is compared with the average value. If it is less than the average value, it is recorded as 0; otherwise, it is recorded as 1. A 256-bit fingerprint of the original image is obtained. The threshold frequency range is 20-30Hz. The filter is a Gaussian filter.

[0020] Furthermore, the grayscale conversion of the color image is achieved using any one of the component method, maximum value method, average value method, or weighted average method.

[0021] In a second aspect, an image supervision method based on privacy intersection and blockchain is implemented using an image supervision platform based on consortium blockchain and privacy intersection. The method includes the following steps:

[0022] (1) The two parties involved in the supervision negotiate and share a key, and use the shared key to encrypt the relevant parameters of the privacy intersection before transmitting them to the other party;

[0023] (2) The parties involved in the supervision construct their own Bloom filters using the image fingerprint data they possess based on the decrypted parameters; specifically, the regulated parties, namely social platform operators and NFT trading platforms, construct obfuscated Bloom filters based on Cuckoo Hash, while the regulatory agencies construct ordinary Bloom filters.

[0024] (3) Both parties involved in the supervision execute an unintentional transmission protocol. The regulated party constructs an element pair for each element in the cuckoo hash-based obfuscated bloom filter and sends the element pair to the regulatory agency. The regulatory agency selects an element from the element pair based on its existing ordinary bloom filters and puts it into a new obfuscated bloom filter.

[0025] (4) The regulatory agency calculates the intersection between its own confidential image library and the user image library of the regulated entity based on the new obfuscated Bloom filter, thereby determining whether the regulated entity possesses illegal images.

[0026] Further, step (2) constructs a cuckoo hash-based obfuscated Bloom filter, which specifically includes the following sub-steps:

[0027] (2.1) First, generate k sub-fingerprints of the original fingerprint, and calculate the corresponding positions of the k sub-fingerprints of the original fingerprint in the confusion Bloom filter according to the k hash functions;

[0028] (2.2) Next, based on the k positions, the k sub-fingerprints are respectively placed into the corresponding positions in the confusion Bloom filter; the k sub-fingerprints fp of a certain image j The relationship between the image and the original fingerprint fp is as follows: j∈(1, 2, ..., k);

[0029] (2.3) If the position for placing the sub-fingerprint is empty, then directly place the sub-fingerprint fp. j Place it in the corresponding position of the confusion Bloom filter;

[0030] (2.4) If the position where the sub-fingerprint is to be placed already contains an element, then place the sub-fingerprint at that position first, and simultaneously use m Cuckoo Hash functions CH in sequence. i Locate a new empty position in the obfuscated Bloom filter for the element that was previously at that position. If no collision occurs while using the Cuckoo Hash function to locate the new position, stop the location operation and place the element to be located in the empty position. Otherwise, place the element to be located in the collision position and continue to use another Cuckoo Hash function to find an empty position for the element that was previously in the collision position. If a collision still occurs after all m Cuckoo Hash functions have been used, then use the last element to be located to overwrite the element in the last located position.

[0031] Furthermore, the mathematical expression for generating the k sub-fingerprints of the original fingerprint is as follows:

[0032]

[0033] Where, r i fp represents a randomly generated string of the same length as the original fingerprint fp. j Let represent the j-th sub-fingerprint, and k%2 indicate whether k is odd. If it is odd, then k%2 equals 1, otherwise it equals 0, and j∈(1,2,...k).

[0034] Furthermore, the execution of the unintentional transmission protocol by both parties involved in the supervision in step (3) specifically includes the following sub-steps:

[0035] (3.1) The regulatory agency initiates an unintentional transfer protocol request to the regulated party, namely the NFT trading platform or social platform operator;

[0036] (3.2) The regulated entity uses a Cuckoo Hash-based Obfuscated Bloom Filter (CBF) based on the populated data. X Generate an element pair (0, CBF) for each position t of the Bloom filter. X [t]);

[0037] (3.3) The regulated party and the regulatory agency execute a 2-to-1 inadvertent transmission protocol bit by bit, wherein the regulated party sends a pre-constructed element pair (0, CBF) to the regulatory agency each time. X [t]), bits represent the number of bits in the Bloom filter;

[0038] (3.4) Regulatory agencies, based on the constructed general filter (BF) Y The 0 or 1 character placed at the corresponding position is used to select an element from the element pair sent by the supervisor, and the selected element is placed in the corresponding position t of a new obfuscated Bloom filter.

[0039] Further, step (3.2) specifically involves the regulated entity using a cuckoo hash-based obfuscated Bloom filter (CBF). X The Bloom filter is traversed starting from the highest bit; when a certain position t is reached, the monitored element constructs an element pair: (0, CBF) X [t]), this element pair consists of 0 and the image sub-fingerprint (CBF) stored in the current position t of the Bloom filter. X [t]) constitutes the structure.

[0040] Further, step (3.4) specifically involves determining the value of position t (BF) in the regulator's Bloom filter. Y When [t]) is 0, the regulator chooses 0 sent by the regulated party, while when the value of t at the corresponding position in the Bloom filter (BF) is 0, the regulator chooses 0. Y When [t]) is 1, the regulator selects the CBF from the element pair sent by the regulated entity. X [t], after selecting an element each time, it is put into a new obfuscated Bloom filter (GBF). Z The formula for the method by which the supervisor selects the corresponding element in the element pair is as follows: (The text appears to be incomplete and contains several errors. A more accurate translation would require the full context.)

[0041]

[0042] Furthermore, step (4) includes the following sub-steps:

[0043] (4.1) The regulatory agency selects the fingerprint (fpx) of one of the images uploaded by its confidentiality agency, and calculates k sub-fingerprints using k hash functions and then applies them to a GBF (Browser-Based Bloom) filter. Z k positions in the array, with k set to 3;

[0044] (4.2) Calculate the k sub-fingerprints of the image in step (1) in GBF using m cuckoo hashes. Z The other m positions in the Bloom filter are obfuscated; each sub-fingerprint has m+1 candidate positions, and m is set to 2 to effectively reduce the probability of hash collisions in the Bloom filter;

[0045] (4.3) Based on the k sub-fingerprints of the current image and the m+1 candidate positions of each sub-fingerprint, calculate all (m+1) k There are k fingerprint combinations; each fingerprint combination has k obfuscation Bloom filters (GBF). Z The string stored in;

[0046] (4.4) For the above (m+1) k Given a set of string combinations, calculate the XOR value of the k strings for each combination, and represent the result as fpy, using the following formula:

[0047]

[0048] Where, p i This indicates the index position of the i-th string in the current string combination within the obfuscated Bloom filter, GBF Z [p i This indicates a blomber filter p that is being obfuscated. i Image fingerprint string at location;

[0049] (4.5) For each round of calculation, the regulator needs to calculate (m+1). k Each fingerprint string is used to calculate the Hamming distance between itself and the image fingerprints owned by the regulator, in order to obtain a minimum Hamming distance D. min The formula is as follows:

[0050]

[0051] Where l represents the length of the image fingerprint string set to 256, fpx[i] represents the 0 and 1 characters of the i-th position of the fingerprint string of the confidential image possessed by the supervisor, and fpy j This indicates that the image that the current regulator wants to calculate is the one using the new obfuscated Bloom filter GBF. Z The fingerprint string composed of the j-th seed fingerprint combination in fpy j [i] represents the 0 or 1 character at the i-th position of the fingerprint string, fpx[i]∈{0,1}, fpy j [i]∈{0,1},1≤j≤(m+1) k ;

[0052] (4.6) Set a Hamming distance threshold D th And compare it with the calculated minimum Hamming distance D of the image fingerprint. min Compare; if D min ≤D tg If the result is positive, it means that the user's image contains a confidential image owned by a classified agency; otherwise, it means that the user's image does not contain a classified image.

[0053] The beneficial effects of this invention are as follows:

[0054] 1. This invention achieves the goal of effectively regulating user-owned private images held by NFT trading platforms and social media platform operators. It ensures that neither the regulatory body nor the regulated platform can access image data beyond their shared area, thus guaranteeing the privacy of user-owned and confidential images.

[0055] 2. This invention uses a consortium blockchain to bring various confidentiality agencies and regulatory agencies under unified management, enabling secure and efficient sharing of fingerprint data of confidential images on the blockchain. This facilitates intersection calculations for regulatory agencies using the confidential image library, while the immutability of the blockchain prevents the image data uploaded by confidentiality agencies from being tampered with.

[0056] 3. The privacy intersection method proposed in this invention improves upon the traditional obfuscated Bloom filter by using Cuckoo Hash to reduce the impact of hash collisions on privacy intersection, thereby improving the accuracy of privacy intersection calculation. Furthermore, compared with traditional time-consuming public-key encrypted privacy intersection schemes, this invention still maintains a significant efficiency advantage. Attached Figure Description

[0057] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0058] Figure 1 This is a flowchart of the image supervision platform based on privacy intersection and consortium blockchain of the present invention;

[0059] Figure 2 This is a flowchart of the core steps of the privacy intersection protocol proposed in this invention;

[0060] Figure 3 This invention uses a confusion Bloom filter (GBF) to store fingerprint data of images;

[0061] Figure 4 This invention uses a cuckoo hash-based obfuscated Bloom filter (CBF) to store image fingerprint data held by regulators;

[0062] Figure 5 This invention uses a common Bloom filter (BF) to store confidential image fingerprint data owned by regulatory agencies;

[0063] Figure 6 This is a flowchart illustrating the process by which the two parties involved in the supervision of this invention transmit image fingerprint data via a 2-choose-1 unintentional transmission protocol.

[0064] Figure 7This is a diagram illustrating the process of element transfer and selection between the regulated entity and the regulatory agency in this invention. Detailed Implementation

[0065] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention as detailed in the appended claims.

[0066] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The singular forms “a,” “the,” and “the” used in this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0067] To address the challenges of data silos, this invention utilizes a consortium blockchain to integrate various institutions holding and managing classified image data into a regulatory system. This enables secure sharing of confidential image data, thereby reducing the difficulty and improving the efficiency of regulation.

[0068] Furthermore, addressing the shortcomings of existing privacy intersection techniques based on traditional obfuscated Bloom filters and unintentional transmissions, this invention proposes an improved Bloom filter based on Cuckoo Hash and combines it with unintentional transmissions to achieve privacy intersection of image datasets. Specifically, by introducing the idea of ​​Cuckoo Hash into the Bloom filter, the accuracy of privacy intersection can be greatly improved, making the solution of privacy intersections more efficient.

[0069] like Figure 1 As shown, an image regulation platform based on the intersection of consortium blockchain and privacy consists of eight core components: consortium blockchain, confidentiality agency, regulatory agency, image owner, social platform operator, image storage system, NFT trading platform, and off-chain storage system.

[0070] The consortium blockchain is a blockchain with strict access control, jointly maintained by confidential and regulatory agencies. Only confidential and regulatory agencies are allowed to read, write, and send transactions on the data in the consortium blockchain.

[0071] The confidential organization is one that possesses a source image library that is prohibited from being publicly disclosed. This organization calculates fingerprints for its own confidential images using an image fingerprinting method, and then uploads them to the consortium blockchain. This allows the image fingerprints to be shared with other organizations on the consortium blockchain, but the original image corresponding to the image fingerprint is not disclosed, and the on-chain data is immutable. It should be noted that the data uploaded to the consortium blockchain here is not the original image data, but the image fingerprint data. Each image fingerprint is only 256 bits long, and the data volume is very small. Therefore, it is feasible to store it on the blockchain as metadata.

[0072] The regulatory body could be multiple national regulatory agencies, used to oversee the legality of personal images uploaded by users to domestic NFT trading platforms and social media operators. The sensitive image dataset used to detect user images originates from the sensitive image fingerprint set uploaded to the blockchain by the aforementioned confidentiality agency.

[0073] The image owners are either ordinary users of social media platforms or users who want to convert their images into NFTs. For social media platform users, image owners can share and publish their private images on the platform, and these images are stored in the image storage system by the social media platform operator. For NFT users, image owners can upload their private images to an NFT trading platform and put the ownership metadata of the image on the blockchain, while the actual image is stored in an off-chain storage system.

[0074] The social platform operator is used to operate an online social platform, providing users with social services such as sending and sharing personal pictures, and user picture data is stored in a picture storage system.

[0075] The image storage system can be a cloud storage system or an offline distributed storage system, used to store personal social media images uploaded by users.

[0076] The NFT trading platform is used to provide NFT trading services to users, receive image data uploaded by users that is to be uploaded to the blockchain, and store it in an off-chain storage system.

[0077] The off-chain storage system is a decentralized storage system used to store the user's original image data before the NFT is uploaded to the blockchain; the decentralized storage system uses the InterPlanetary File System (IPFS), which disperses the user's original image data across different storage nodes in a decentralized manner.

[0078] The NFT trading platform retrieves user-uploaded images from an off-chain storage system and calculates the corresponding fingerprints using an image fingerprinting method. Similarly, the social media platform operator retrieves user-shared social media images from an image storage system and then calculates the fingerprints of the corresponding images.

[0079] The method for calculating the image fingerprint can be a traditional hash algorithm, but this invention preferably uses an average hash algorithm. The implementation steps of this algorithm can be described as follows: First, the color image can be converted to grayscale using the component method, maximum value method, average value method, or weighted average method. Then, a filter is used to perform noise reduction processing on the image; the filter can be a Gaussian filter, a statistical median filter, or a bilateral filter. Next, discrete wavelet transform technology is used to transform the image, ultimately obtaining a low-frequency image. In digital image processing, the commonly used low-frequency threshold is between 20 and 30 Hz. When the frequency components of an image are below this threshold, this part of the image can be considered a low-frequency image. Finally, based on the low-frequency image, it is reduced to a size of 16×16, and the average pixel grayscale of the low-frequency image is calculated. Then, each pixel is compared with the average value; if it is less than the average value, it is recorded as 0; otherwise, it is recorded as 1. Thus, a 256-bit fingerprint of the original image is obtained.

[0080] The traditional hash algorithm used in this invention can be any one of perceptual hashing, differential hashing, average hashing, and wavelet hashing. The perceptual hashing algorithm converts the image into a set of digital features through DCT transformation, then compresses these features into a fixed-length binary string to obtain the image's fingerprint information. The differential hashing algorithm obtains the image's fingerprint information by calculating the grayscale difference between adjacent pixels. The average hashing algorithm scales the image to a fixed size, converts it into a grayscale image, calculates its average grayscale value, and then compares the grayscale value of each pixel with the average grayscale value, marking pixels with higher grayscale values ​​as 1 and others as 0, thus obtaining the image's fingerprint information. The wavelet hashing algorithm uses wavelet transform to decompose the image into multiple frequency bands, then performs hash calculations on each frequency band to finally obtain the image's fingerprint information.

[0081] Through the steps described above, NFT trading platforms and social media operators obtain the fingerprints corresponding to private images uploaded or shared by users. Not all of these images are legal; users may use NFT trading platforms and social media platforms to share and disseminate illegal images. Therefore, regulatory agencies need to regulate these NFT trading platforms and social media operators to prevent the spread and sharing of illegal images. It is important to note that NFT regulation must be conducted before the images are listed on the blockchain as digital assets. If regulatory agencies intervene after an image has been publicly circulated on the blockchain as an NFT, they will lose the right to prohibit the public display of the image.

[0082] The regulatory body reads image fingerprints uploaded by various confidential organizations from the consortium blockchain. These fingerprints may not all originate from a single confidential organization or may come from different confidential organizations on the consortium blockchain. After the regulatory body has read a certain batch of image fingerprints from the blockchain, it can begin executing a privacy intersection protocol with the NFT trading platform and social platform operators to resolve the intersection of image data between the regulatory body and the image owners.

[0083] It is important to emphasize that when regulatory agencies oversee user images, they must not disclose users' private images. In other words, regulatory platforms need to conduct image oversight without obtaining users' private and private images. More importantly, in interactions with NFT trading platforms and social media operators, regulatory agencies must also not disclose image data uploaded to the consortium blockchain by confidential organizations.

[0084] To achieve the above objectives, this invention proposes an image supervision method based on privacy intersection. By solving the privacy intersection problem, it not only enables regulatory agencies to detect whether users possess illegal images, but also ensures that neither party involved in the privacy intersection, namely the regulatory agency and the user's image data holding agency, will disclose the images they possess.

[0085] like Figure 2 As shown, the two-party privacy intersection protocol proposed in this invention consists of four core stages:

[0086] (1) The two parties involved in the supervision negotiate to share a key, and use the shared key to encrypt the relevant parameters of the privacy request before transmitting them to the other party.

[0087] (2) The parties involved in the regulation construct their respective Bloom filters using the image fingerprint data they possess based on the decrypted parameters. Specifically, the regulated parties, namely social platform operators and NFT trading platforms, construct obfuscated Bloom filters based on Cuckoo Hash, while the regulatory agencies construct ordinary Bloom filters.

[0088] (3) Both parties involved in the supervision execute an unintentional transmission protocol. The regulated party constructs an element pair for each element in the Bloom filter and sends the element pair to the regulatory agency. The regulatory agency selects an element from the element pair based on its Bloom filters and puts it into a new obfuscated Bloom filter.

[0089] (4) The regulatory agency calculates the intersection between its own confidential image library and the user image library of the regulated entity based on the new obfuscated Bloom filter.

[0090] In the first phase of the privacy intersection protocol, to ensure that both parties involved in the oversight can share the privacy intersection parameters and guarantee the security of the privacy intersection, this invention first uses the Diffie-Hellman key exchange algorithm to negotiate the key used to encrypt the privacy intersection-related parameters. Next, the overseer selects the number of bits in the Bloom filter and the hash function, then encrypts the parameters using the specified key and sends them to the overseen party, i.e., the NFT trading platform or social platform operator. Upon receiving the encrypted privacy intersection parameters, the overseen party uses the specified key to decrypt the ciphertext parameters, obtaining the plaintext parameters.

[0091] In the second phase of the privacy request protocol, regulators and regulated entities can construct their respective Bloom filters based on the parameters to prepare for subsequent privacy requests.

[0092] The Bloom filters mentioned include two types: one is the traditional Bloom filter (BF) that only stores 0 and 1 characters, and the other is the Garbled Bloom filter (GBF) that stores arbitrary strings. Both can be used in privacy-preserving intersection schemes. Figure 3 As shown, the principle of a Bloom filter is to use k hash functions to obtain the k positions in the Bloom filter that an element will be mapped to, and then place the k components of the element into the corresponding positions. Similarly, when searching for whether an element appears in a set of elements, all elements in the set can first be mapped to the Bloom filter using the k hash functions of the Bloom filter. Then, for the element to be searched, its k positions to be mapped are searched again. If the components at the k positions are all equal to the components at the corresponding k positions in the Bloom filter, then the element is in the set. Figure 3 In this context, fp represents the element to be mapped, specifically image fingerprint data, while the fp stored in the Bloom filter... j (j∈(1,2,...k)) represents the k components of the original fingerprint of the image, referred to in this invention as the k sub-fingerprints of the image. It is important to emphasize that the traditional obfuscated Bloom filter (GBF) is not a simple Bloom filter that only stores 0 and 1 characters. The obfuscated Bloom filter is an improved Bloom filter where each position stores a binary string of a specified size. This invention specifies and uses a 256-bit string of 0s and 1s.

[0093] This invention proposes a novel Bloom filter to address the shortcomings of traditional Cuckoo Hash-based Bloom filters: the Cuckoo Hash-based Bloom Filter (CBF). This CBF utilizes the Cuckoo Hash (CH) concept to reduce hash collisions to some extent, thereby improving the accuracy of image data intersection calculation. Furthermore, this improved method is used in the image monitoring scheme of this invention to enhance the quality of monitoring.

[0094] like Figure 4 As shown, the present invention employs a special processing method for mapping the k sub-fingerprints of an image to a confused Bloom filter based on cuckoo hashing. Referring to the idea of ​​cuckoo hashing, the cuckoo hashing function is used again at the location where a collision occurs to map the image sub-fingerprint data at that location to find a new unoccupied index location, thereby reducing the probability of hash collisions to a certain extent.

[0095] More specifically, in Figure 4 In the image fingerprint fp, the first sub-fingerprint fp1 is originally mapped to position 8 in the Bloom filter. Since the 8th bit of the Bloom filter is empty, the final position of sub-fingerprint fp1 is also 8. For the second sub-fingerprint fp2, its original mapping position in the Bloom filter is 2, but this second position already contains element fp. x At this point, fp2 is still stored in the second position, but the original fp in that position... x The element is evicted and stored in a new position 4 after being mapped by the first Cuckoo Hash function CH1. Since position 4 is an empty position in the Bloom filter, there is no need to move fp. x Remapping, 4 is used as fp x The final position. Similarly, for the k-th sub-fingerprint fp of fingerprint fp. k Its original mapping position in the Bloom filter is 5, so 5 is fp. k The final position. However, element fp already exists in position 5. y It will be mapped to the new position 7 through the first cuckoo hash function CH1, while the original element fp in position 7 will be... w It will be mapped to position 9 through the second cuckoo hash function CH2. Since position 9 is initially empty, 9 is fp. w The final position.

[0096] It should be noted that the hash functions used to resolve collisions in the cuckoo hash-based obfuscated Bloom filter described above in this invention are not the k hash functions used in the Bloom filter, but rather m additionally defined cuckoo hash functions. Since their number is also customizable, a larger number generally results in a lower probability of hash collisions. This is because if a collision still occurs after m cuckoo hashes, the old element at the collision location will be overwritten by the new element, which affects the accuracy of privacy intersection. To reduce the impact on privacy intersection efficiency, the number m of cuckoo hash functions is set to 2. Of course, in specific implementations, this value can be customized as needed; however, setting it too large will affect the efficiency of privacy intersection. Therefore, when setting the number of cuckoo hash functions, a balance needs to be struck between the accuracy and efficiency of privacy intersection.

[0097] The above-mentioned cuckoo hash-based obfuscated Bloom filter first calculates the positions of k sub-fingerprints in the Bloom filter using k hash functions. Then, based on these k positions, the k sub-fingerprints are placed into their corresponding positions within the cuckoo hash-based obfuscated Bloom filter. The k sub-fingerprints fp of a certain image... j The relationship between the image and the original fingerprint fp is shown in formula (1):

[0098]

[0099] More specifically, for each fingerprint component fp j These are all constructed based on fingerprint fp. To ensure that each fingerprint component is associated with the original fingerprint as much as possible, and that the XOR value of each fingerprint component equals the original fingerprint, this invention proposes a new construction method, as shown in the following formula:

[0100]

[0101] In formula (2), r i This represents a randomly generated string of the same length as fp. k%2 indicates whether k is odd; if odd, k%2 equals 1, otherwise it equals 0. It should be noted that regardless of whether k is odd or even, the k sub-fingerprints fp constructed according to the above formula... j The XOR result is the same as the original fingerprint fp.

[0102] According to the above method, the NFT trading platform and the social platform operator will sequentially solve the corresponding fingerprints of the user's private images using the image fingerprint calculation method, and put the sub-fingerprint of each user image into a specified position in a cuckoo hash-based obfuscated Bloom filter.

[0103] Similarly, such as Figure 5As shown, the regulatory agency also needs to save the confidential image data uploaded by the confidential organization from the consortium blockchain into a Bloom filter. However, unlike NFT trading platforms and social platform operators, the regulatory agency uses a regular Bloom filter that only stores 0 and 1 characters. Although the regulatory agency does not store the sub-fingerprints of the images in the Bloom filter, in order to match the positions with the cuckoo hash-based obfuscated Bloom filter to complete the intersection with the privacy image data of NFT trading platforms and social platform operators, the regulatory agency uses k hash functions to map the fingerprint data, and simultaneously uses the same number of m cuckoo hash functions to set all m hash positions of the image sub-fingerprint data to 1.

[0104] Following the steps outlined above, both the regulatory body and the regulated entities—the NFT trading platforms and social media operators—store their respective image fingerprint data in Bloom filters. In other words, each party possesses two pre-filled Bloom filters. More specifically, the NFT trading platforms and social media operators possess a confused Bloom filter generated using Cuckoo Hash, which stores the sub-fingerprint data for each image. The regulatory body, on the other hand, possesses a standard Bloom filter, which stores binary characters of 0s and 1s.

[0105] After constructing the appropriate Bloom filter, the two parties involved in the oversight process enter the third phase of the privacy request protocol, namely, the execution of the Oblivious Transfer (OT) protocol. For example... Figure 6 As shown, the detailed steps for executing the unintentional transport protocol are as follows:

[0106] S1: The regulatory agency initiates an unintentional transfer protocol request to the regulated party, namely the NFT trading platform or social platform operator.

[0107] S2: The regulated entity uses a Cuckoo Hash-based Obfuscated Bloom Filter (CBF) based on the populated data. X Generate an element pair for each location of the Bloom filter. Specifically, the regulated entity obtains an element pair from the CBF. X The Bloom filter is traversed starting from the highest bit. When a certain position t is reached, the monitored element constructs an element pair: (0, CBF) X [t]), this element pair consists of 0 and the image sub-fingerprint (CBF) stored in the current position t of the Bloom filter. X [t]) constitutes the structure.

[0108] S3: The regulated entity and the regulatory entity execute a 2-to-1 inadvertent transmission protocol bit by bit. Each time, the regulated entity sends a pre-constructed element pair (0, CBF) to the regulatory entity. X [t]).

[0109] S4: Regulatory agencies based on the constructed general filter (BF) Y The 0 or 1 character placed at the corresponding position is used to select an element from the element pair sent by the supervisor, and the selected element is placed in the corresponding position t of a new obfuscated Bloom filter.

[0110] In step 4, the method by which the regulator selects the corresponding element in the element pair can be described by formula (3), when the value of the corresponding position t in the regulator's Bloom filter (BF) Y When [t]) is 0, the regulator chooses 0 sent by the regulated party, while when the value of t at the corresponding position in the Bloom filter (BF) is 0, the regulator chooses 0. Y When [t]) is 1, the regulator selects the CBF from the element pair sent by the regulated entity. X [t]. For example, Figure 7 As shown, each time the regulator selects an element, it is placed into a new Confusion Bloom Filter (CBF). Z The corresponding position t in ) .

[0111]

[0112] It should be noted that this new obfuscated Bloom filter (GBF) constructed by the regulatory authorities... Z It is also a Bloom filter that can store strings of a specified size, unlike ordinary Bloom filters that can only store 0 and 1 characters.

[0113] Finally, after receiving all the data from the Bloom filter sent by the regulated entity, the regulatory agency begins the fourth phase of the privacy intersection protocol, which involves calculating the intersection of the image data. The specific process is as follows:

[0114] The regulatory agency selects the fingerprint (fpx) of one of the images uploaded by its confidential agency, calculates k sub-fingerprints using k hash functions, and then applies them to a GBF obfuscation Bloom filter. Z The k positions in the equation. In this invention, to balance privacy, intersection efficiency, and error rate, k is uniformly set to 3.

[0115] Next, using m cuckoo hashes, k sub-fingerprints of the image are calculated in GBF. Z The other m positions in the Bloom filter are obfuscated. Therefore, each sub-fingerprint has m+1 candidate positions. In this invention, m is set to 2 to effectively reduce the probability of hash collisions in the Bloom filter.

[0116] Based on the k sub-fingerprints of the current image and the m+1 candidate positions of each sub-fingerprint, calculate all (m+1) k There are k fingerprint combinations. Each fingerprint combination has k obfuscation Bloom filters (GBF).Z The string stored in it.

[0117] For the above (m+1) k For each string combination, the XOR value of the k strings is calculated as shown in formula (4), and the result is represented as fpy. Where p i This indicates the index position of the i-th string in the current string combination within the obfuscated Bloom filter, GBF Z [p i This indicates a blomber filter p that is being obfuscated. i The image fingerprint string at the location.

[0118]

[0119] After solving for fpy, the regulatory agency can directly compare its size with the image fingerprint fpx in its confidential image library to determine whether the confidential image appears in the user's image library. In order to more accurately identify similar images, this invention introduces Hamming distance to measure the similarity between two images. It should be further noted that the Hamming distance is not calculated directly on the original pixels of the two images, but is calculated based on the fingerprint data of the two images. As shown in formula (5), l represents the length of the image fingerprint string, which is uniformly set to 256 in this invention. D(fpx, fpy) represents the Hamming distance between fingerprint fpx and fingerprint fpy, fpx[i] represents the 0 and 1 characters of the i-th position of the fingerprint string of the confidential image owned by the regulator, and similarly, fpy[i] represents the 0 and 1 characters of the i-th position of the fingerprint string calculated by the regulator according to the new obfuscated Bloom filter, fpx[i]∈{0,1}, fpy[i]∈{0,1}.

[0120]

[0121] As shown in formula (6), for each round of calculation, the regulator needs to calculate (m+1). k Each fingerprint string is used to calculate the Hamming distance between itself and the image fingerprints owned by the regulator, in order to obtain a minimum Hamming distance D. min In formula (6), fpy j , 1≤j≤(m+1) k This indicates that the image that the current regulator wants to calculate is the one using the new obfuscated Bloom filter GBF. Z The fingerprint string is composed of the j-th seed fingerprint combination.

[0122]

[0123] It is important to further emphasize that after introducing cuckoo hashing into the Bloom filter, when comparing image fingerprints, regulatory agencies not only need to calculate the XOR value fpy of the original sub-fingerprints in the newly obfuscated Bloom filter mapped to k positions by the k hash functions of the Bloom filter, but also need to calculate a total of (m+1) k Seed fingerprint combination. The reason why (m+1) appears is... k The seed fingerprint combination method is because the k positions mapped by the k hash functions of the Bloom filter are not fixed. The original sub-fingerprints in these k positions may be replaced by m cuckoo hash functions in other m positions.

[0124] Finally, by setting a Hamming distance threshold D th And compare it with the calculated minimum Hamming distance D of the image fingerprint. min Compare. If D min ≤D th If the result is positive, it means that the user's image contains a confidential image owned by a classified agency; otherwise, it means that the user's image does not contain a classified image.

[0125] It should be noted that the Hamming distance threshold D is... tg The threshold can be set according to the actual situation. Generally, the smaller the Hamming distance threshold, the more likely it is that an illegal image will be identified in the user's image library only when the user's image is highly similar to the confidential image. However, if the image has undergone some image processing, it will be impossible to accurately identify it.

[0126] Following the steps outlined above, regulators can determine whether NFT trading platforms and social media platform operators have stored illegally uploaded images by sequentially examining the image fingerprints they possess from the consortium blockchain. If the regulated entity possesses illegal images, the regulatory agency can notify the relevant platform to provide information about the users involved for legal action.

[0127] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. The invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the following claims.

[0128] It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. An image supervision platform based on the intersection of consortium blockchain and privacy, characterized in that, The platform consists of eight parts: consortium blockchain, image owners, social platform operators, image storage system, NFT trading platform, off-chain storage system, confidentiality agency, and regulatory agency. The consortium blockchain is a blockchain with strict access control jointly maintained by a confidentiality agency and a regulatory agency. Only the confidentiality agency and the regulatory agency are allowed to read, write, and send transactions on the data in the consortium blockchain. The owners of the images are ordinary users of social media platforms or users who have a need to convert their image works into NFTs; The social platform operator is used to operate an online social platform and provide users with social services such as sending and sharing personal pictures; it obtains social pictures shared by users from the image storage system and then calculates the fingerprint of the corresponding picture according to the image fingerprint calculation method; The image storage system is a cloud storage system or an offline distributed storage system, used to store personal social images uploaded by users; The NFT trading platform is used to provide NFT trading services to users, receive image assets uploaded by users that are to be put on the blockchain, and store them in the off-chain storage system; retrieve the images uploaded by users from the off-chain storage system, and then calculate the fingerprint of the corresponding image according to the image fingerprint calculation method; The off-chain storage system is a decentralized storage system used to store the user's original image data before the NFT is uploaded to the blockchain; the decentralized storage system uses the InterPlanetary File System (IPFS), which decentralizes and distributes the user's original image data across different storage nodes. The confidential organization is an organization that possesses a source image library that is prohibited from being disclosed. This organization calculates the fingerprint of its own confidential images according to the image fingerprint calculation method, and then uploads it to the consortium blockchain, so that the fingerprint of the image is shared with other organizations on the consortium blockchain, but the original image corresponding to the image fingerprint is not disclosed, and the data on the chain is immutable. The regulatory bodies are several regulatory bodies used to supervise the legality of personal images uploaded by users to NFT trading platforms and social platform operators; to read image fingerprints uploaded by confidential institutions from the consortium blockchain, where the image fingerprints come from the same confidential institution or from different confidential institutions on the consortium blockchain; and to supervise NFT trading platforms and social platform operators. The image owner uploads the image to the social platform operator or NFT trading platform; the social platform operator stores the user-uploaded images in a distributed image storage system, and the NFT trading platform stores the user-uploaded images to be uploaded to the blockchain in the off-chain storage system; the confidentiality agency calculates the fingerprints of all images in its confidential image library and uploads these image fingerprints to the consortium blockchain; The social platform operator or NFT trading platform acts as the regulated party, and the regulatory agency acts as the regulator. The two parties communicate with each other to negotiate relevant parameters for privacy requests. The regulated party obtains user-uploaded images in batches from the image storage system and off-chain storage system, calculates the fingerprints of these images, and then constructs a confusion Bloom filter based on Cuckoo Hash. The regulatory body obtains fingerprint datasets of confidential images uploaded by the various confidential organizations from the consortium blockchain and constructs ordinary filters based on these fingerprint data. The two parties involved in the supervision execute an unintentional transmission protocol based on their respective Bloom filters, and the regulator determines whether the regulated party possesses illegal images uploaded by the user based on the results of the unintentional transmission.

2. The image monitoring platform based on the intersection of consortium blockchain and privacy as described in claim 1, characterized in that, The image fingerprint calculation method is any one of the perceptual hash algorithm, differential hash algorithm, average hash algorithm, and wavelet hash algorithm; The average hash algorithm is as follows: First, the color image is converted to grayscale. Then, a filter is used to perform noise reduction on the image, and a discrete wavelet transform is used to transform the image to obtain an image with a frequency below a threshold. Finally, based on the obtained image with a frequency below the threshold, it is reduced to a size of 16×16, and the average pixel grayscale of the image with a frequency below the threshold is calculated. Then, each pixel is compared with the average value. If it is less than the average value, it is recorded as 0; otherwise, it is recorded as 1. A 256-bit fingerprint of the original image is obtained. The threshold frequency range is 20~30 Hz. The filter is a Gaussian filter.

3. The image monitoring platform based on the intersection of consortium blockchain and privacy as described in claim 2, characterized in that, The process of converting a color image to grayscale is achieved using any one of the following methods: component method, maximum value method, average value method, or weighted average method.

4. An image supervision method based on privacy intersection and blockchain, implemented on an image supervision platform based on consortium blockchain and privacy intersection, characterized in that... The method includes the following steps: (1) The two parties involved in the supervision negotiate and share a key, and use the shared key to encrypt the relevant parameters of the privacy intersection before transmitting them to the other party; (2) The parties involved in the supervision construct their own Bloom filters using the image fingerprint data they possess based on the decrypted parameters; specifically, the regulated parties, namely social platform operators and NFT trading platforms, construct obfuscated Bloom filters based on Cuckoo Hash, while the regulatory agencies construct ordinary Bloom filters. (3) Both parties involved in the supervision execute an unintentional transmission protocol. The regulated party constructs an element pair for each element in the cuckoo hash-based obfuscated bloom filter and sends the element pair to the regulatory agency. The regulatory agency selects an element from the element pair based on its existing ordinary bloom filters and puts it into a new obfuscated bloom filter. (4) The regulatory agency calculates the intersection between its own confidential image library and the user image library of the regulated entity based on the new obfuscated Bloom filter, thereby determining whether the regulated entity possesses illegal images.

5. The image supervision method based on privacy intersection and blockchain according to claim 4, characterized in that, Step (2) constructs a cuckoo hash-based obfuscated Bloom filter, which specifically includes the following sub-steps: (2.1) First, generate k sub-fingerprints of the original fingerprint, and calculate the corresponding positions of the k sub-fingerprints of the original fingerprint in the confusion Bloom filter according to the k hash functions; (2.2) Next, based on the k positions, the k sub-fingerprints are placed into the corresponding positions in the confusion Bloom filter; let the k sub-fingerprints of a certain image... Original fingerprint of the image The relationship is: , ; (2.3) If the position for placing the sub-fingerprint is empty, then directly place the sub-fingerprint. Place it in the corresponding position of the confusion Bloom filter; (2.4) If the position where the sub-fingerprint is to be placed already contains an element, then place the sub-fingerprint at that position first, and simultaneously use m Cuckoo Hash functions in sequence. Locate a new empty position in the obfuscated Bloom filter for the element that was previously at that position. If no collision occurs while using the Cuckoo Hash function to locate the new position, stop the location operation and place the element to be located in the empty position. Otherwise, place the element to be located in the collision position and continue to use another Cuckoo Hash function to find an empty position for the element that was previously in the collision position. If a collision still occurs after all m Cuckoo Hash functions have been used, then use the last element to be located to overwrite the element in the last located position.

6. The image supervision method based on privacy intersection and blockchain according to claim 5, characterized in that, The mathematical expression for generating the k sub-fingerprints of the original fingerprint is as follows: ; in, The length of the randomly generated fingerprint is compared to the original fingerprint. The same string, Indicates the first Fingerprint This indicates whether k is odd; if it is odd, then... It equals 1, otherwise it equals 0. .

7. The image supervision method based on privacy intersection and blockchain according to claim 4, characterized in that, In step (3), the parties involved in the supervision execute the unintentional transmission protocol, which specifically includes the following sub-steps: (3.1) The regulatory agency initiates an unintentional transfer protocol request to the regulated party, namely the NFT trading platform or social platform operator; (3.2) The regulated entity uses a cuckoo hash-based obfuscated Bloom filter on the populated data. For each location of the Bloom filter Generate an element pair (0, ); (3.3) The implementation of the regulations by the regulated parties and the regulatory agencies This is a 2-out-of-1 unintentional transmission protocol, where the regulated party sends a pre-constructed pair of elements (0, 1) to the regulatory agency each time. ), Indicates the number of bits in the Bloom filter; (3.4) Regulatory agencies, based on the constructed general filters ( The 0 or 1 character placed at the corresponding position is used to select a specific element from the element pair sent by the supervisor, and the selected element is placed in the corresponding position of a new obfuscated Bloom filter. middle.

8. The image supervision method based on privacy intersection and blockchain according to claim 7, characterized in that, Specifically, step (3.2) involves the regulated entity using a cuckoo hash-based obfuscated Bloom filter. The high-order part begins to traverse the Bloom filter; When traversing to a certain position At that time, the regulated entity constructs an element pair: (0, This element pair consists of 0 and the image sub-fingerprint stored in the current position t of the Bloom filter. It is composed of ).

9. The image supervision method based on privacy intersection and blockchain according to claim 7, characterized in that, Specifically, step (3.4) involves the corresponding position in the regulator's Bloom filter... The value ( When the value is 0, the regulator selects 0 sent by the regulated party, while when the corresponding position in the Bloom filter... The value ( When ) is 1, the regulator selects the element pair sent by the regulated entity. Each time an element is selected, it is put into a new confusion Bloom filter. The corresponding position in ) The formula for the regulator to select the corresponding element in the element pair is as follows: 。 10. The image supervision method based on privacy intersection and blockchain according to claim 4, characterized in that, Step (4) includes the following sub-steps: (4.1) The fingerprint of one of the images uploaded by the regulatory agency that it possesses for confidentiality purposes. The k sub-fingerprints are calculated using k hash functions and then processed by a Bloom filter. k positions in the array, with k set to 3; (4.2) Calculate the k sub-fingerprints of the image in step (1) using m cuckoo hashes. The other m positions in the Bloom filter are obfuscated; each sub-fingerprint has m+1 candidate positions, and m is set to 2 to effectively reduce the probability of hash collisions in the Bloom filter; (4.3) Based on the k sub-fingerprints of the current image and the m+1 candidate positions of each sub-fingerprint, calculate all... There are k fingerprint combinations; each fingerprint combination has k obfuscation Bloom filters. The string stored in; (4.4) Regarding the above Given a set of k string combinations, for each combination, calculate the XOR value of the k strings, and represent the result as follows: The formula is as follows: ; in, This indicates the index position of the i-th string in the current string combination within the obfuscation Bloom filter. This indicates a blomber filter with obfuscation. Image fingerprint string at location; (4.5) For each round of calculation, the regulator needs to calculate Each fingerprint string is used to calculate the Hamming distance between itself and the image fingerprints owned by the regulator, in order to obtain a minimum Hamming distance value. The formula is as follows: ; in, This indicates that the length of the image fingerprint string is set to 256. The 0 and 1 characters in the i-th position of the fingerprint string representing the confidential image possessed by the regulator. This indicates that the image that the current regulator wants to calculate is the one using the new obfuscated Bloom filter. The fingerprint string is composed of the j-th seed fingerprint combination. This represents the 0 or 1 character at the i-th position of the fingerprint string. , , ; (4.6) Set a Hamming distance threshold And compare it with the calculated minimum Hamming distance of the image fingerprint. Compare; if If the result is positive, it means that the user's image contains a confidential image owned by a classified agency; otherwise, it means that the user's image does not contain a classified image.

Citation Information

Patent Citations

  • Private data intersection calculation method and device, equipment and storage medium

    CN113343305A

  • Multi-party privacy set intersection method, system and device

    CN114726611A

  • High-efficiency extensible safe document similarity calculating method and device

    CN107885705A

  • Pathogenic gene detection method based on privacy protection intersection calculation protocol

    CN111125736A