Copyright tracking verification method for digital collections

Abstract

The invention provides a copyright tracking verification method for a digital collection, and relates to the technical field of data processing, and the method comprises the steps: generating a first digital fingerprint; receiving copyright holding information and creation information of the target digital collection, and uploading the copyright holding information and creation information together with the first digital fingerprint to a block chain network; when the first user uploads purchased collections to the block chain network for copyright authentication, generating a second digital fingerprint and proof information; performing validity verification, and when the validity verification is passed, generating a collection matching result; if the collection matching result passes, copyright tracking verification is carried out based on copyright holding information in the registration result, and a copyright tracking verification result is generated. Through the copyright authentication method and device, the technical problems of privacy leakage and data security risk caused by insufficient privacy protection of the collection uploaded by the user in the prior art can be solved, and the technical effects of improving the data security during copyright authentication and ensuring the reliability of copyright authentication are achieved.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method for copyright tracking and verification of digital collections. Background Technology

[0002] With the rapid development of the digital economy, digital collectibles have attracted widespread attention due to their unique value attributes and convenient circulation methods. However, copyright authentication and tracking issues are becoming increasingly prominent in the process of digital collectible transactions. In recent years, blockchain and artificial intelligence technologies have been gradually applied to the field of copyright protection for digital collectibles. Blockchain technology, with its decentralized, tamper-proof, and traceable characteristics, provides a new solution for the copyright authentication and tracking of digital collectibles.

[0003] Most existing digital artifact copyright authentication methods rely on blockchain to match the feature fingerprints of digital artifacts to achieve copyright authentication. However, users may face privacy leaks and data security risks when uploading digital artifacts for authentication.

[0004] In summary, existing technologies suffer from insufficient privacy protection for user-uploaded collections, leading to risks of privacy leaks and data security breaches. Summary of the Invention

[0005] The purpose of this application is to provide a copyright tracking and verification method for digital collectibles, in order to solve the technical problems in the prior art that lead to privacy leaks and data security risks due to insufficient privacy protection for user-uploaded collectibles.

[0006] In view of the above problems, this application provides a copyright tracking and verification method for digital collectibles, comprising: extracting multimodal features from the original data of the target digital collectible using a multimodal generative adversarial network to generate a first digital fingerprint; receiving the copyright holding information and creation information of the target digital collectible, and uploading them together with the first digital fingerprint to a blockchain network for copyright registration, generating a registration result, wherein the copyright holding information includes information on holders of multiple rights; when a first user uploads a purchased collectible to the blockchain network for copyright authentication, combining an encryption module with the multimodal generative adversarial network to extract features and encrypt the purchased collectible, generating a second digital fingerprint and proof information; the first user receiving the proof information and verifying its validity through the encryption module, and when the validity verification is successful, matching the first digital fingerprint and the second digital fingerprint to generate a collectible matching result; if the collectible matching result is successful, performing copyright tracking verification based on the copyright holding information in the registration result to generate a copyright tracking verification result.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: A first digital fingerprint is generated by extracting multimodal features from the raw data of the target digital collectible using a multimodal generative adversarial network (GAN). The copyright holding information and creation information of the target digital collectible are received and uploaded to the blockchain network along with the first digital fingerprint for copyright registration, generating a registration result. The copyright holding information includes information on holders of various rights. When a first user uploads a purchased collectible to the blockchain network for copyright authentication, the purchased collectible is feature-extracted and encrypted using an encryption module and the GAN, generating a second digital fingerprint and proof information. The first user receives the proof information and verifies its validity using the encryption module. If the validity verification passes, the first and second digital fingerprints are matched to generate a collectible matching result. If the collectible matching result passes, copyright tracking verification is performed based on the copyright holding information in the registration result, generating a copyright tracking verification result. By using a multimodal generative adversarial network to extract multimodal features from the raw data of the target digital collectibles and establish a first digital fingerprint, copyright registration is carried out on the blockchain network in combination with copyright holding information and creation information. When a user uploads a purchased collectible for copyright authentication, feature extraction and encryption are performed in combination with an encryption module to match the collectibles, thereby improving the data security during copyright authentication and ensuring the reliability of copyright authentication.

[0008] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in this application 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 merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0010] Figure 1 This is a flowchart illustrating a copyright tracking and verification method for digital collectibles according to this application; Figure 2 This is a schematic diagram of the process for generating a first digital fingerprint in a copyright tracking and verification method for digital collectibles according to this application. Detailed Implementation

[0011] This application provides a copyright tracking and verification method for digital collectibles, addressing the technical problem in existing technologies where insufficient privacy protection for user-uploaded collectibles leads to privacy leaks and data security risks. By using a multimodal generative adversarial network (GAN) to extract multimodal features from the raw data of the target digital collectible and establish a first digital fingerprint, this fingerprint is combined with copyright holding and creation information for copyright registration on a blockchain network. When a user uploads a purchased collectible for copyright authentication, an encryption module is used for feature extraction and encryption to match the collectible results, thereby improving data security during copyright authentication and ensuring its reliability.

[0012] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0013] Please see the appendix Figure 1 This application provides a method for copyright tracking and verification of digital collectibles, which specifically includes the following steps: Step 1: Extract multimodal features from the original data of the target digital collection using a multimodal generative adversarial network to generate the first digital fingerprint.

[0014] Specifically, a multimodal generative adversarial network (GAN) is a deep learning model specifically designed to process and synthesize data with multiple data types or modalities. A target digital artifact refers to a specific digital asset requiring feature extraction and identification, such as a unique digital artwork or collectible. Multimodal feature extraction refers to the process by which a GAN identifies and extracts key features from the raw data of the target digital artifact. For example, for a digital painting, the network might extract visual features such as color distribution, texture patterns, and shape contours, as well as semantic features from any textual descriptions associated with the painting. For digital artifacts containing audio, the network might also extract auditory features such as frequency distribution and rhythmic patterns.

[0015] The generated first digital fingerprint is a unique digital identifier created based on these extracted multimodal features. This fingerprint reflects the unique attributes and characteristics of the digital collection and can be used to verify its authenticity and uniqueness. Technically, this fingerprint may be a multidimensional vector or code, where each dimension represents a feature extracted from different modalities of data.

[0016] Step 2: Receive the copyright holding information and creation information of the target digital collectible, and upload them together with the first digital fingerprint to the blockchain network for copyright registration, generating a registration result. The copyright holding information includes information on holders of various rights.

[0017] Specifically, copyright holding information typically includes information about the holders of various rights, such as the original author's name, contact information, copyright statement, and information about potential authorized users, collaborators, or successors. Creation information involves the time and place of creation, the background of creation, and the technology or tools used.

[0018] Copyright holding information, creation information, and the first digital fingerprint are uploaded to the blockchain network. Blockchain, a distributed database technology, is an ideal choice for digital copyright management due to its decentralized, immutable, and transparent characteristics. Registering copyright on the blockchain means permanently recording copyright information and the digital fingerprint, ensuring the immutability and traceability of copyright information. Simultaneously, to protect the privacy of copyright holders and ensure information security, the relevant information can be encrypted before being uploaded to the blockchain network. The encrypted copyright holding information, creation information, and digital fingerprint are packaged into a transaction and broadcast to the blockchain network. Nodes in the blockchain network verify and confirm this transaction according to a predetermined consensus mechanism (such as proof-of-work or proof-of-stake). Once the transaction is confirmed by network consensus, the relevant information is recorded in a new block and permanently added to the blockchain. After copyright registration is completed, a registration result is generated, usually a unique registration number or certificate, proving that the copyright information of the digital collectible has been registered on the blockchain. In this way, any copyright inquiry, transaction, or dispute resolution can be conducted through the blockchain, greatly improving the efficiency and transparency of copyright management.

[0019] Step 3: When the first user uploads the purchased collectible to the blockchain network for copyright authentication, the encryption module and the multimodal generative adversarial network are used to extract and encrypt the features of the purchased collectible, generating a second digital fingerprint and proof information.

[0020] Specifically, when the first user—the person who purchased the target digital collectible—wishes copyright authentication for their purchase, the request is processed using a combination of an encryption module and a multimodal generative adversarial network (GAN). First, the GAN extracts multimodal features from the purchased digital collectible, similar to the process of generating the first digital fingerprint. The GAN analyzes the collectible's image, text, and audio data, extracting multimodal features such as color, texture, shape, and semantic content. These extracted features are then encrypted using an encryption module. Encryption ensures the security of the uploaded collectible information during transmission and storage, preventing unauthorized access and data leakage. The generated second digital fingerprint, created based on the encrypted features, is a unique identifier used to verify the authenticity and ownership of the purchased collectible. Simultaneously, proof information is generated, obtained using zero-knowledge proof technology. Zero-knowledge proof is an encryption protocol that allows one party to prove the truth of a statement to another without revealing any additional information, enhancing the security of the entire authentication process and protecting user privacy.

[0021] Finally, the second digital fingerprint and proof information are uploaded to the blockchain network. The blockchain network verifies this information and compares it with the previously registered first digital fingerprint. If the comparison results match, the copyright authentication is successful, and the user's ownership is confirmed and recorded on the blockchain. This process not only protects the rights of copyright holders but also provides buyers with guarantees of ownership confirmation and anti-counterfeiting.

[0022] Step 4: The first user receives the proof information and verifies its validity through the encryption module. When the validity verification is successful, the first digital fingerprint and the second digital fingerprint are matched to generate a collection matching result.

[0023] Specifically, after receiving the proof information, the first user needs to verify its validity through an encryption module, specifically using a zero-knowledge proof protocol within the encryption module. Once the proof information passes validity verification, the next step is to match the first and second digital fingerprints. This step is to confirm that the collectible purchased by the user matches the copyright information registered on the blockchain, i.e., verifying the authenticity and copyright ownership of the collectible. Specifically, the second digital fingerprint is compared with the first digital fingerprint recorded on the blockchain. If they match perfectly, it means that the collectible held by the user is authentic and the copyright information has been registered. A collectible matching result is generated based on the fingerprint comparison result. If the fingerprint matches successfully, a matching success message is displayed; if the fingerprint does not match, a matching failure message is displayed, alerting the user to potential risks.

[0024] Through this series of verification and matching processes, users can be assured of the authenticity and copyright ownership of the digital collectibles they purchase, thereby protecting users' rights and enhancing the trust level of the digital collectibles market.

[0025] Step 5: If the collection matching result is successful, perform copyright tracking verification based on the copyright holding information in the registration result, and generate a copyright tracking verification result.

[0026] Specifically, once the matching result of the collectibles is successful—that is, the first digital fingerprint matches the second digital fingerprint—it is confirmed that the digital collectibles held by the user are authentic and the copyright is registered. Copyright tracking and verification will then be conducted based on the copyright holding information in the registration result. This step is to further confirm the copyright transfer path and the current copyright holding status, ensuring the legality and transparency of copyright transactions. First, copyright holding information related to the target digital collectible is extracted from the blockchain network. This information includes detailed information about the original author, authorized users, collaborators, or successors—all copyright-related parties. Further, the source of the purchased collectible is obtained, and it is determined whether the source matches the copyright holding information. A copyright tracking and verification result is generated to confirm the legality of the copyright, helping to prevent copyright infringement and illegal transactions.

[0027] Further details are attached. Figure 2 As shown, step one of this application includes: The process involves: acquiring the collection attributes of the target digital collection and the preset extraction modal features of the multimodal generative adversarial network; constructing a feature generation sample set based on the collection attributes and the preset extraction modal features; training the multimodal generative adversarial network based on the feature generation sample set; extracting multimodal features from the original data of the target digital collection through the multimodal generative adversarial network; and performing digitization processing based on the extraction results to generate the first digital fingerprint.

[0028] Specifically, the steps for generating the first digital fingerprint include: firstly, acquiring the collection attributes and pre-defined extraction modal features. Collection attributes include the type of collection (e.g., artwork, music, video, etc.). Pre-defined extraction modal features refer to the types of features the multimodal generative adversarial network is designed to extract during training, such as visual features, audio features, or text features. These are determined by experts in the field based on collection attributes and practical experience. For example, collection samples can be obtained based on collection attributes, and features from different dimensions can be extracted for collection differentiation and recognition. The feature with the highest recognition accuracy is then used as the pre-defined extraction modal features.

[0029] Next, a feature generation sample set is constructed based on the collection attributes and preset extracted modal features. This sample set is used to train a multimodal generative adversarial network (GAN) to extract multimodal features related to the collection attributes from the raw data. The process of constructing the sample set may include selecting data similar to the collection attributes and preset extracted modal features from an existing collection database, or using data augmentation techniques to expand the sample set.

[0030] Then, the pre-constructed feature generation sample set is used to train the multimodal generative adversarial network (GAN). During training, the GAN learns how to extract pre-defined modal features from the input raw data. Through continuous learning and adjustment, it improves its ability to extract features from multimodal data. After training, the GAN is used to extract multimodal features from the raw data of the target digital artifacts. That is, the trained GAN is used to analyze the raw data of the target artifacts, such as images, audio, and text, and extract features from them.

[0031] Finally, the extracted results are digitized to generate the first digital fingerprint. Digitization involves converting the extracted features into a digital fingerprint. This digital fingerprint is a unique identifier for the target artifact, containing key features extracted from its multimodal data. Through this process, the multimodal features of the target digital artifact are effectively extracted and encoded into a digital fingerprint, providing a technological foundation for subsequent copyright authentication and tracking.

[0032] Furthermore, this application also includes the following steps: The multimodal generative adversarial network is connected to the hash calculation module. The hash calculation module receives the output of the multimodal generative adversarial network and performs hash value calculation to generate the first digital fingerprint. The hash calculation module includes a hash algorithm.

[0033] Specifically, the multimodal generative adversarial network (GAN) is connected to the hash calculation module, working together to generate the first digital fingerprint. The GAN is responsible for extracting features from the multimodal raw data of the target digital artifact, while the hash calculation module calculates hash values ​​for these features to generate the digital fingerprint.

[0034] The hash calculation module includes a hash algorithm, which converts input data into a fixed-length string. A key characteristic of hash algorithms is that even minute changes in the input will result in completely different outputs, ensuring a high degree of uniqueness in the hash value. Even minor differences in the collection data can be reflected in the digital fingerprint. The hash calculation module first receives the output of a multimodal generative adversarial network (GAN), i.e., the features extracted from the multimodal data of the target digital collection. To adapt to the input requirements of the hash algorithm, the features extracted by the GAN need to undergo certain transformations or encodings. For example, the feature vectors are converted into a data format suitable for hash algorithm processing. The hash algorithm is then used to calculate the hash value from the transformed features. Commonly used hash algorithms include SHA-256 and MD5. The generated hash value is the first digital fingerprint. This fingerprint is a fixed-length string representing the unique characteristics of the target digital collection.

[0035] In this way, the combination of multimodal generative adversarial networks and hash calculation modules can not only extract rich features from multimodal data, but also convert these features into unique and secure digital fingerprints for copyright authentication and anti-counterfeiting of digital collectibles.

[0036] Furthermore, this application also includes the following steps: The multimodal generative adversarial network (GAN) includes a generative neural network (GNN), a discriminative neural network (DNN), and an adversarial training layer. The GNN extracts multimodal features from input samples in the feature generation sample set based on the preset modal features. The DNN discriminates the feature extraction results output by the GNN based on output samples in the feature generation sample set and outputs a feature loss metric. The adversarial training layer optimizes the network parameters of the GNN based on the feature loss metric, iterating multiple times until the GNN converges.

[0037] Specifically, a multimodal generative adversarial network (GAN) comprises three main parts: a generative neural network (GNN), a discriminative neural network (DNN), and an adversarial training layer. These parts work together, through adversarial training, enabling the GNN to effectively extract multimodal features from input samples. The GNN extracts multimodal features from the input samples in the feature generation sample set based on predefined modal features. Its role is to simulate the distribution of real multimodal data and generate multimodal features that are as close as possible to real data. Predefined modal features refer to the types of features that the network should focus on and extract, such as visual, audio, or textual features, determined during network design.

[0038] The discriminative neural network (DNN) is responsible for judging the feature extraction results of the generative neural network (GAN) based on the output samples (which can also be understood as real samples) in the feature generation sample set. The goal is to distinguish between the output samples and the feature extraction results from the GAN, i.e., to determine whether the output samples and the feature extraction results from the GAN are consistent. The discriminative neural network output feature loss metric measures the difference between the feature extraction results from the GAN and the output samples. The adversarial training layer optimizes the network parameters of the GAN based on the discriminative neural network output feature loss metric. Its purpose is to iteratively optimize the GAN's parameters so that it can generate features that increasingly closely resemble the output samples, making it difficult for the discriminative neural network to distinguish between the output samples and the feature extraction results from the GAN. This process iterates multiple times until the GAN converges, meaning its feature extraction results are sufficiently close to the output samples that the discriminative neural network cannot effectively distinguish. Through this adversarial training method, multimodal generative adversarial networks (GANs) can effectively extract features from multimodal data and generate a first digital fingerprint for copyright authentication and anti-counterfeiting of digital collectibles.

[0039] Furthermore, step four of this application includes: The encryption module includes a homomorphic encryption unit and a zero-knowledge proof unit. The first user encrypts the data of the purchased collection using the homomorphic encryption algorithm within the homomorphic encryption unit to generate ciphertext. The ciphertext is then digitized after multimodal feature analysis by the multimodal generative adversarial network to generate a second digital fingerprint. A proof information is then created for the ciphertext using the zero-knowledge proof protocol within the zero-knowledge proof unit. The proof information includes a proof question and a proof result.

[0040] Specifically, the steps for extracting and encrypting features of the purchased collectibles using the encryption module and the multimodal generative adversarial network to generate a second digital fingerprint and proof information are as follows: The encryption module includes a homomorphic encryption unit and a zero-knowledge proof unit. These two units work together to provide security for the copyright authentication of digital collectibles. The homomorphic encryption unit includes a homomorphic encryption algorithm. Homomorphic encryption is a special encryption method that allows users to perform calculations on encrypted data without decryption first. This allows users to perform operations on encrypted data while maintaining data privacy. The homomorphic encryption algorithm within the homomorphic encryption unit is used to encrypt the collectible data purchased by the first user, generating ciphertext of the collectibles. Thus, even in an encrypted state, necessary processing and analysis of the collectible data can still be performed. The specific homomorphic encryption algorithm is selected by those skilled in the art from existing technologies.

[0041] Multimodal generative adversarial networks (GANs) are used to perform multimodal feature analysis on the encrypted ciphertext of the collection, extracting features from it. This step is performed under encryption to ensure the privacy and security of the collection data. Based on the features extracted by the GANs, a second digital fingerprint is generated using the aforementioned hash calculation module. This fingerprint is a unique identifier used to verify the authenticity and ownership of the purchased collection.

[0042] The zero-knowledge proof unit includes a zero-knowledge proof protocol, an encryption protocol that allows one party to prove the truth of a statement to another without revealing any additional information. The zero-knowledge proof protocol within the unit is used to create proof information for the encrypted data of a collection. This proof information includes a proof question and a proof result. For example, if the encrypted data of a collection is x, the prover chooses a random number y and calculates h = H(x,y) using the zero-knowledge proof protocol, where H is the proof algorithm in the protocol, such as a hash function. y is the proof question, and h is the proof result. Subsequently, the first user calculates H(x,y) and compares the result with h. If they are equal, the prover knows x without revealing its specific value. In this way, the encryption module not only protects the user's data privacy but also ensures the security of the copyright authentication process.

[0043] Furthermore, this application also includes the following steps: The first user receives the proof information and performs verification calculations on the proof problem using the verification algorithm defined in the zero-knowledge proof protocol, generating a verification calculation result; it then determines whether the verification calculation result is consistent with the proof result. If so, the validity verification passes, and a consistency comparison is performed between the first digital fingerprint and the second digital fingerprint to generate the collection matching result.

[0044] Specifically, after receiving the proof information, the first user needs to verify the proof problem using the verification algorithm defined within the zero-knowledge proof protocol to ensure the authenticity and validity of the proof information. The first user uses the verification algorithm defined in the zero-knowledge proof protocol to calculate the proof problem. This algorithm is designed based on the principles of the zero-knowledge proof protocol. The specific zero-knowledge proof protocol is determined by professionals in the field based on existing technologies, such as ZK-SNARKs (Zero-Knowledge Concise Non-Interactive Knowledge Proofs) and ZK-STARKs (Zero-Knowledge Concise Transparent Extensible Proofs), which can verify the validity of the proof information without revealing any detailed information about the collection. Through the calculation of the verification algorithm, the first user obtains a verification calculation result. This result is generated based on the proof problem and is used to compare with the proof result. Then, it is determined whether the verification calculation result is consistent with the proof result. If they are consistent, it means that the proof information is valid, the statements in the proof information are true, and the validity verification passes.

[0045] Next, the first user performs a consistency comparison between the first and second digital fingerprints to confirm that the purchased collectible matches the copyright information registered on the blockchain. If the second digital fingerprint is a perfect match with the first, it indicates that the collectible held by the user is authentic and the copyright information has been registered, generating a collectible matching result. In this way, the zero-knowledge proof protocol and fingerprint comparison jointly ensure the security, validity, and accuracy of digital collectible copyright authentication, providing users with a reliable copyright authentication mechanism.

[0046] Furthermore, step five of this application includes: Read the provenance information of the purchased collectibles; compare the copyright type and copyright holder based on the provenance information and the copyright holding information, and generate the copyright tracking verification result.

[0047] Specifically, the process involves retrieving the provenance information of purchased collectibles. This provenance information typically includes historical ownership records, purchase channels, and transaction history, reflecting the collection's journey from its original creator to its current owner. The provenance information is then compared with copyright holding information. This information includes detailed details about the original creator, authorized users, collaborators, or successors. The comparison process verifies whether the copyright holder in the provenance information matches the copyright holding information registered on the blockchain. If the copyright holder information matches, the copyright tracking verification passes; if discrepancies exist, the verification fails, and the specific inconsistencies are provided. Through this comparison and verification process, users and copyright holders can be assured of the legality of the copyright and the transparency of the transfer path, helping to prevent copyright infringement and illegal transactions, and protecting the rights of copyright holders.

[0048] Furthermore, this application also includes the following steps: A collection tracking and monitoring network is set up based on the multimodal features of the target digital collection; wherein, the collection tracking and monitoring network is connected to an Internet media port, and when the collection tracking and monitoring network detects a collection whose feature similarity with the multimodal features is greater than a preset similarity, a second reminder message is generated.

[0049] Specifically, based on the multimodal characteristics of the target digital collectible, a collectible tracking and monitoring network can be set up. The main function of this network is to monitor internet media ports to detect whether other digital collectibles have a feature similarity exceeding a preset similarity threshold with the target collectible. If a similar collectible is detected, a second alert message will be generated.

[0050] The collection tracking and monitoring network connects to internet media ports to monitor digital collections on the internet in real time. When the network detects that the characteristics of other digital collections on the internet are more similar to those of the target collection than a preset similarity threshold, it generates a second alert. This second alert is sent to the copyright holder or relevant stakeholders, reminding them of potential copyright infringement or illegal copying. In this way, the collection tracking and monitoring network provides effective monitoring and protection for digital collections.

[0051] In summary, the copyright tracking and verification method for digital collectibles provided in this application has the following technical effects: A first digital fingerprint is generated by extracting multimodal features from the raw data of the target digital collectible using a multimodal generative adversarial network (GAN). The copyright holding information and creation information of the target digital collectible are received and uploaded to the blockchain network along with the first digital fingerprint for copyright registration, generating a registration result. The copyright holding information includes information on holders of various rights. When a first user uploads a purchased collectible to the blockchain network for copyright authentication, the purchased collectible is feature-extracted and encrypted using an encryption module and the GAN, generating a second digital fingerprint and proof information. The first user receives the proof information and verifies its validity using the encryption module. If the validity verification passes, the first and second digital fingerprints are matched to generate a collectible matching result. If the collectible matching result passes, copyright tracking verification is performed based on the copyright holding information in the registration result, generating a copyright tracking verification result. By using a multimodal generative adversarial network to extract multimodal features from the raw data of the target digital collectibles and establish a first digital fingerprint, copyright registration is carried out on the blockchain network in combination with copyright holding information and creation information. When a user uploads a purchased collectible for copyright authentication, feature extraction and encryption are performed in combination with an encryption module to match the collectibles, thereby improving the data security during copyright authentication and ensuring the reliability of copyright authentication.

[0052] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0053] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for copyright tracking and verification of digital collectibles, characterized in that, include: Multimodal features are extracted from the raw data of the target digital collection using a multimodal generative adversarial network to generate the first digital fingerprint; The copyright holding information and creation information of the target digital collectible are received and uploaded to the blockchain network together with the first digital fingerprint for copyright registration, generating a registration result. The copyright holding information includes information on holders of various rights. When the first user uploads the purchased collectibles to the blockchain network for copyright authentication, the encryption module and the multimodal generative adversarial network are combined to extract features and encrypt the purchased collectibles, generating a second digital fingerprint and proof information. The first user receives the proof information and verifies its validity through the encryption module. When the validity verification is successful, the first digital fingerprint and the second digital fingerprint are matched to generate a collection matching result. If the collection is matched, copyright tracking and verification are performed based on the copyright holding information in the registration result, and a copyright tracking and verification result is generated.

2. The copyright tracking and verification method for digital collectibles as described in claim 1, characterized in that, Multimodal feature extraction is performed on the raw data of the target digital artifact using a multimodal generative adversarial network to generate a first digital fingerprint, including: Obtain the collection attributes of the target digital collection and the preset extracted modal features of the multimodal generative adversarial network; A feature generation sample set is constructed based on the collection attributes and the preset extracted modal features; The multimodal generative adversarial network is trained based on the sample set generated from the aforementioned features; The multimodal generative adversarial network is used to extract multimodal features from the original data of the target digital collection, and the extraction results are then digitized to generate the first digital fingerprint.

3. The copyright tracking and verification method for digital collectibles as described in claim 2, characterized in that, The multimodal generative adversarial network is connected to the hash calculation module. The hash calculation module receives the output of the multimodal generative adversarial network and performs hash value calculation to generate the first digital fingerprint. The hash calculation module includes a hash algorithm.

4. The copyright tracking and verification method for digital collectibles as described in claim 3, characterized in that, The multimodal generative adversarial network includes a generative neural network, a discriminative neural network, and an adversarial training layer; The generative neural network extracts multimodal features from the input samples in the feature generation sample set based on the preset extraction modal features, and the discriminative neural network discriminates the feature extraction results output by the generative neural network based on the output samples in the feature generation sample set, and outputs a feature loss index. The adversarial training layer optimizes the network parameters of the generative neural network based on the feature loss metric, iterating multiple times until the generative neural network converges.

5. The copyright tracking and verification method for digital collectibles as described in claim 1, characterized in that, The encryption module, in conjunction with the multimodal generative adversarial network, performs feature extraction and encryption on the purchased collectibles to generate a second digital fingerprint and proof information, including: The encryption module includes a homomorphic encryption unit and a zero-knowledge proof unit; The first user encrypts the data of the purchased collection using the homomorphic encryption algorithm within the homomorphic encryption unit, generating ciphertext of the collection. After performing multimodal feature analysis on the encrypted text of the collection through the multimodal generative adversarial network, the text is digitized to generate the second digital fingerprint, and proof information is created on the encrypted text of the collection through the zero-knowledge proof protocol in the zero-knowledge proof unit. The proof information includes the proof question and the proof result.

6. The copyright tracking and verification method for digital collectibles as described in claim 5, characterized in that, The first user receives the zero-knowledge proof and verifies its validity through the encryption module. If the validity verification passes, the first digital fingerprint and the second digital fingerprint are matched to generate a collection matching result, including: The first user receives the proof information and performs verification calculations on the proof problem using the verification algorithm defined within the zero-knowledge proof protocol, generating a verification calculation result; Determine whether the verification calculation result is consistent with the proof result. If so, the validity verification is passed. Perform a consistency comparison between the first digital fingerprint and the second digital fingerprint to generate the collection matching result.

7. The copyright tracking and verification method for digital collectibles as described in claim 1, characterized in that, Based on the copyright holding information in the registration results, copyright tracking and verification are performed to generate copyright tracking and verification results, including: Read the provenance information of the purchased collectibles; Based on the information on the provenance of the collection and the information on copyright ownership, the copyright type and copyright holder are compared to generate the copyright tracking and verification result.

8. The copyright tracking and verification method for digital collectibles as described in claim 1, characterized in that, Also includes: A collection tracking and monitoring network is set up based on the multimodal characteristics of the target digital collection; The collection tracking and monitoring network is connected to an internet media port. When the collection tracking and monitoring network detects a collection whose feature similarity to the multimodal features is greater than a preset similarity, it generates a second reminder message.

Images