AI digital content copyright storage method and device based on block chain and medium

Abstract

The invention discloses an AI digital content copyright evidence storage method and device based on a block chain and a medium, and relates to the technical field of copyright authentication. The method comprises the following steps: acquiring to-be-authenticated data, and storing the to-be-authenticated data in a block chain; wherein the to-be-authenticated data comprises text data, image data and audio and video data; inputting the to-be-authenticated data into a preset multi-modal feature extraction model to generate target authentication data; based on a preset smart contract, performing ownership verification on the target authentication data, and generating a verification result; and correspondingly writing the target authentication data passing the authentication and the verification result into the block chain. Different types of data are quickly processed through the multi-modal feature extraction model, and the smart contract automatically executes verification operation, so that the efficiency of AI multi-modal digital content copyright evidence storage is improved.

Description

Technical Field

[0001] The present application relates to the field of copyright authentication technology, and in particular to a blockchain-based AI digital content copyright notarization method, device, and medium. Background Art

[0002] In today's digital age, multimodal large models are flourishing. They enable the collaborative generation of digital content such as text, images, audio, and video, greatly improving content production efficiency and enriching the digital content ecosystem. However, with the rapid development of AI-powered multimodal digital content generation technology, abuses such as deepfakes and voice impersonation are becoming increasingly rampant, seriously threatening the security and stability of the content ecosystem.

[0003] Traditional blockchain-based evidence storage solutions exhibit significant limitations when dealing with multimodal feature fusion and massive amounts of data. For example, a single-chain structure can only support text hash storage, making it difficult to preserve key ownership identifiers such as video watermarks and voiceprints. Furthermore, as the number of nodes increases, evidence storage latency increases linearly, impacting efficiency.

[0004] Therefore, how to preserve copyright for AI multimodal digital content has become a technical issue that needs to be solved urgently. Summary of the Invention

[0005] The embodiments of the present application provide a blockchain-based AI digital content copyright notarization method, device, and medium to solve the technical problem of how to perform copyright notarization for AI multimodal digital content.

[0006] In the first aspect, an embodiment of the present application provides an AI digital content copyright notarization method based on blockchain, the method comprising: obtaining data to be authenticated and storing the data to be authenticated on the blockchain; wherein the data to be authenticated includes text data, image data, audio and video data; inputting the data to be authenticated into a preset multimodal feature extraction model to generate target authentication data; based on a preset smart contract, performing ownership verification on the target authentication data and generating a verification result; and writing the authenticated target authentication data and the verification result into the blockchain accordingly.

[0007] In one implementation of the present application, the data to be authenticated is input into a preset multimodal feature extraction model to generate target authentication data, specifically including: extracting features from the data to be authenticated to obtain a multimodal feature vector; based on a cross-modal attention mechanism, dynamically weighting and aggregating the multimodal feature vectors using a gated fusion network to generate a joint feature vector; and performing a hierarchical hash operation on the joint feature vector to obtain the target authentication data.

[0008] In one implementation of the present application, feature extraction is performed on the authentication data to obtain a multimodal feature vector, specifically including: natural language feature encoding of text data to generate a text feature vector; visual feature encoding of image data to generate an image feature vector; spatiotemporal feature encoding of audio and video data to generate an audio and video joint feature vector; and correlating the text feature vector, image feature vector and audio and video joint feature vector to obtain a multimodal feature vector.

[0009] In one implementation of the present application, based on the cross-modal attention mechanism, a gated fusion network is used to dynamically weight and aggregate multimodal feature vectors to generate a joint feature vector, specifically including: based on the cross-modal attention mechanism, calculating the cross-modal attention score between the text feature vector, the image feature vector and the audio and video joint feature vector, and generating a weight coefficient corresponding to the cross-modal attention score; inputting the weight coefficient into the gated fusion network, and dynamically correcting the weight coefficient based on the gating mechanism preset in the gated fusion network; based on the corrected weight coefficient, performing weighted summation on the multimodal feature vectors to generate a joint feature vector.

[0010] In one implementation of the present application, based on a preset smart contract, the target authentication data is verified for ownership and a verification result is generated, specifically including: based on a preset main chain-side chain cross-chain evidence network transmission protocol, the target authentication data is sent from the side chain to the main chain; according to the verification consensus rules of the main chain, a basic verification is performed on the format integrity and data source legitimacy of the target authentication data; after the basic verification is passed, the preset smart contract in the main chain is mobilized to send the verification process data back to the side chain through the cross-chain evidence network transmission protocol; based on the multi-factor authentication mechanism of the side chain, the verification process is analyzed, and the ownership of the target authentication data is judged to generate a verification result.

[0011] In one implementation of the present application, a multi-factor authentication mechanism based on a side chain is used to analyze the verification process and determine the ownership of the target authentication data to generate a verification result, specifically including: decrypting the digital signature based on the creator's public key in the target authentication data to generate a signature validity identifier; extracting the biometric hash value in the target authentication data, and performing similarity calculation with the registration features pre-stored in the off-chain security database to generate a biometric matching identifier; accessing the external copyright registration system through the on-chain oracle module to query the off-chain ownership status based on the content fingerprint of the target authentication data and generate an ownership registration validity identifier; if the signature validity identifier, biometric matching identifier and ownership registration validity identifier all return a pass status, a valid ownership verification result is generated and the evidence storage on-chain operation is triggered; otherwise, an ownership exception alarm result containing a failure factor type is generated.

[0012] In one implementation of the present application, after generating the verification result, the method also includes: if an ownership anomaly alarm result is received, generating a derivative work identifier and a corresponding derivative work based on the content fingerprint in the target authentication data; storing the derivative work identifier on the blockchain, and sending the derivative work to the copyright detection node and conducting a secondary copyright review.

[0013] In one implementation of the present application, the authenticated target authentication data and the verification result are written into the blockchain correspondingly, specifically including: extracting the content fingerprint of the target authentication data, generating a random confusion factor and calculating the anonymized content hash; encrypting the target authentication data using a symmetric encryption algorithm, and binding it with the anonymized content hash to generate a privacy evidence package; writing the privacy evidence package into the side chain through the cross-chain evidence network, and synchronously writing the anonymized content hash into the main chain.

[0014] On the second aspect, the embodiment of the present application also provides an AI digital content copyright notarization device based on blockchain, the device including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor so that the at least one processor can: obtain data to be authenticated and store the data to be authenticated on the blockchain; wherein the data to be authenticated includes text data, image data, audio and video data; input the data to be authenticated into a preset multimodal feature extraction model to generate target authentication data; based on a preset smart contract, verify the ownership of the target authentication data and generate a verification result; write the authenticated target authentication data and the verification result into the blockchain accordingly.

[0015] On the third aspect, the embodiment of the present application also provides a non-volatile computer storage medium for blockchain-based AI digital content copyright notarization, which stores computer executable instructions, and the computer executable instructions are set to: obtain data to be authenticated and store the data to be authenticated on the blockchain; wherein the data to be authenticated includes text data, image data, audio and video data; input the data to be authenticated into a preset multimodal feature extraction model to generate target authentication data; based on a preset smart contract, verify the ownership of the target authentication data and generate a verification result; write the authenticated target authentication data and the verification result into the blockchain accordingly.

[0016] The embodiments of the present application provide a blockchain-based AI digital content copyright notarization method, device, and medium, which have the following beneficial effects: utilizing the decentralized and tamper-proof characteristics of blockchain to ensure the authenticity and integrity of data; through a preset multimodal feature extraction model, key features can be extracted from different types of data and target authentication data can be generated, and important information in the data can be automatically identified and extracted, avoiding the subjectivity and inaccuracy of manual feature extraction; the multimodal feature extraction model can quickly process different types of data, the smart contract can automatically perform verification operations, and the distributed storage and consensus mechanism of the blockchain can quickly complete data writing, effectively improving the efficiency of AI multimodal digital content copyright notarization. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 A flowchart of a blockchain-based AI digital content copyright notarization method provided in an embodiment of the present application; Figure 2 A schematic diagram of the internal structure of a blockchain-based AI digital content copyright notarization device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0018] To make the purpose, technical solutions, and advantages of this application more clear, the technical solutions of this application will be clearly and completely described below in conjunction with the specific embodiments of this application and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0019] The embodiments of the present application provide a blockchain-based AI digital content copyright notarization method, device, and medium to solve the technical problem of how to perform copyright notarization for AI multimodal digital content.

[0020] The technical solutions proposed in the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0021] Figure 1 This is a flowchart of a blockchain-based AI digital content copyright notarization method provided in the embodiment of this application. Figure 1 As shown, the embodiment of the present application provides an AI digital content copyright notarization method based on blockchain, which specifically includes the following steps: Step 10: Obtain the data to be authenticated and store it on the blockchain.

[0022] In this step, the data to be authenticated can be digital content in various forms, such as text, images, audio, or video. This data may originate from different creators, such as individual creators, corporate users, or content generation platforms. The core characteristics of blockchain are decentralization, immutability, and traceability. Decentralization means there is no single control center, and the security and integrity of the data are maintained by the entire network. Immutability means that once data is written to the blockchain, it cannot be modified or deleted. Traceability means that the source and history of the data can be traced through the blockchain's chain structure. Blockchain is a distributed ledger technology that stores data across multiple nodes using encryption algorithms and consensus mechanisms. Each node maintains a complete copy of the ledger, which is interconnected and synchronized via the network. Storing the data to be authenticated on the blockchain prevents modification or deletion, providing strong guarantees for data security and credibility.

[0023] Step 20: Input the data to be authenticated into a preset multimodal feature extraction model to generate target authentication data.

[0024] As an optional embodiment, the data to be authenticated is input into a preset multimodal feature extraction model to generate target authentication data, which may specifically include: Step 201: extracting features from the data to be authenticated to obtain a multimodal feature vector.

[0025] As an optional embodiment, feature extraction is performed on the authentication data to obtain a multimodal feature vector, which may specifically include: Step 2011: natural language feature encoding is performed on the text data to generate a text feature vector.

[0026] In this step, natural language feature encoding refers to converting the semantic information in text data into numerical feature vectors through specific algorithms or models. The purpose of this process is to express the complex semantics in the text in a form that can be understood and processed by computers. Since text data usually contains rich semantic information, such as the meaning of words, the structure of sentences, the theme of the text, etc., the text data is preprocessed through natural language processing technology, including word segmentation, stop word removal, part-of-speech tagging and other operations, which helps to extract key information from the text and remove irrelevant noise. The preprocessed text data is feature encoded through the BERT language model to capture the semantic information in the text and convert it into a high-dimensional feature vector to facilitate subsequent fusion and analysis with data from other modalities.

[0027] Step 2012: Perform visual feature encoding on the image data to generate an image feature vector.

[0028] In this step, visual feature encoding refers to converting the visual information in the image data into a numerical feature vector through a specific algorithm or model. The purpose of this process is to represent the color, texture, shape, spatial relationship and other information in the image in a form that can be understood and processed by a computer. Since image data usually contains rich visual information, computer vision technology is used to perform normalization, cropping, scaling and other operations on the image data. The pre-processed image data is feature encoded through the ViT visual model to capture the visual information in the image and convert it into a high-dimensional feature vector. The generated image feature vector can reflect the visual content of the image, which is convenient for subsequent fusion and analysis with data from other modalities.

[0029] Step 2013: Perform spatiotemporal feature encoding on the audio and video data to generate an audio and video joint feature vector.

[0030] In this step, spatiotemporal feature encoding refers to converting the time and space information in audio and video data into numerical feature vectors through specific algorithms or models. Since audio and video data not only contains audio signals, but also contains dynamic visual information related to time, it is necessary to consider the characteristics of both time and space dimensions at the same time; the audio signal is denoised, the video signal is extracted from the video frame, and the pre-processed audio and video signal is feature encoded through the Wav2Vec2 audio and video model to capture the time and space features therein, and convert it into a joint audio and video feature vector to reflect the comprehensive content of the audio and video data, which is convenient for subsequent fusion and analysis with other modal data.

[0031] Step 2014: Correlate the text feature vector, the image feature vector, and the audio and video joint feature vector to obtain a multimodal feature vector.

[0032] In this step, the converted text feature vector, image feature vector and audio and video joint feature vector are subjected to multimodal feature fusion to fully utilize the complementary information between different modal data, improve the overall understanding and representation ability of digital content, and provide powerful feature support for subsequent operations such as ownership authentication, infringement detection and content tracing.

[0033] Step 202: Based on the cross-modal attention mechanism, a gated fusion network is used to dynamically weight and aggregate the multimodal feature vectors to generate a joint feature vector.

[0034] As an optional embodiment, based on the cross-modal attention mechanism, a gated fusion network is used to dynamically weight and aggregate multimodal feature vectors to generate a joint feature vector, which may specifically include: Step 2021: Based on the cross-modal attention mechanism, the cross-modal attention score between the text feature vector, the image feature vector and the audio and video joint feature vector is calculated, and a weight coefficient corresponding to the cross-modal attention score is generated.

[0035] In this step, the cross-modal attention mechanism is a method for processing multimodal data. Its purpose is to determine the importance of each modal feature by calculating the relationship between different modal feature vectors. This scheme inputs the text feature vector, image feature vector and audio and video joint feature vector into the cross-modal attention mechanism, evaluates the degree of semantic association between different modal feature vectors, obtains the attention score of each modal feature, and obtains the weight coefficient corresponding to the cross-modal attention score through normalization. The weight coefficient is used in the subsequent feature fusion process to determine the contribution of each modal feature in the fusion result.

[0036] Step 2022: Input the weight coefficient into the gated fusion network, and dynamically modify the weight coefficient based on the gate mechanism preset in the gated fusion network.

[0037] In this step, the gated fusion network is a neural network structure used to dynamically adjust the feature fusion weights. Its core idea is to dynamically adjust the weights of each modal feature according to the characteristics of the current input data through a gating mechanism, thereby achieving more flexible feature fusion. In this scheme, the generated weight coefficients are input into the gated fusion network, and the gating mechanism adjusts the weight coefficients of each modal feature according to the characteristics of the current input data (such as content type, task requirements, etc.). For example, if the current task focuses more on image information, the gating mechanism will increase the image feature weight coefficient and appropriately reduce the weight coefficients of other modal features. In complex multimodal data processing scenarios, this scheme can better capture the complex relationship between modal data by dynamically adjusting the weight coefficients, so that the feature fusion process can better adapt to different scenarios and needs.

[0038] Step 2023: Based on the modified weight coefficients, perform weighted summation on the multimodal feature vectors to generate a joint feature vector.

[0039] In this step, the text feature vector, image feature vector and audio and video joint feature vector are weighted and summed according to the corrected weight coefficients, which can highlight the important features of different data to be authenticated while retaining the information of other features, thereby generating a more representative and adaptable joint feature vector.

[0040] Step 203: Perform a hierarchical hash operation on the joint feature vector to obtain target authentication data.

[0041] In this step, the hierarchical hash operation is a process of gradually simplifying complex data structures into fixed-length hash values. Through multi-layer hash calculations, the joint feature vector is gradually converted into a more concise and unique hash value. This hierarchical processing method can effectively process high-dimensional feature vectors while ensuring that the final generated hash value can uniquely identify the original data. It not only retains the key information of the original data, but also ensures the uniqueness and non-tamperability of the data, providing a reliable foundation for subsequent blockchain storage and data authentication.

[0042] Step 30: Based on the preset smart contract, the target authentication data is verified for ownership and a verification result is generated.

[0043] As an optional embodiment, based on a preset smart contract, the target authentication data is verified for ownership and a verification result is generated, which may specifically include: Step 301: Based on the preset main chain-side chain cross-chain evidence network transmission protocol, the target authentication data is sent from the side chain to the main chain.

[0044] In this step, this solution constructs a main chain-side chain dual-chain evidence storage network. The main chain is responsible for global supervision and storage of core data, while the side chain is used to process specific business data and evidence information; the cross-chain evidence network transmission protocol is a preset communication mechanism for securely and efficiently transmitting data between the main chain and the side chain. It defines the format, security mechanism, verification process and specific rules for cross-chain interoperability of data transmission to ensure that the transmission of data between the two blockchains is reliable and consistent, and to prevent data loss, tampering or errors during transmission.

[0045] Step 302: Based on the verification consensus rules of the main chain, perform basic verification on the format integrity and data source legitimacy of the target authentication data.

[0046] In this step, the verification consensus rules of the main chain are a mechanism for ensuring the authenticity and integrity of the data. The rules define the conditions that the target authentication data must meet before it is accepted and stored. After receiving the target authentication data, the main chain performs a format integrity check on it, including checking whether the data complies with the predefined structure and encoding rules. At the same time, the main chain also verifies whether the source of the target authentication data is legal. This involves checking whether the data generation process complies with the preset rules and whether the data is sent by an authorized side chain node. The legitimacy check can prevent unauthorized data from entering the main chain.

[0047] Step 303: After the basic verification is passed, the preset smart contract in the main chain is mobilized to send the verification process data back to the side chain through the cross-chain evidence network transmission protocol.

[0048] In this step, the smart contract preset in the main chain is specially designed to process the target authentication data. It contains the logic for verifying the authenticity and integrity of the data, as well as the function of recording the verification process and results. When the target authentication data passes the basic verification of the main chain, the smart contract on the main chain will be triggered. The intermediate data and results generated by the smart contract when performing the verification operation record the verification process of the target authentication data, including the verification time, verification steps, verification results, etc., and the verification process data is securely transmitted back to the side chain. This mechanism not only improves the automation and efficiency of copyright authentication, but also ensures the credibility and traceability of the data through the combination of cross-chain protocols and smart contracts, providing solid technical support for the confirmation and protection of digital content.

[0049] Step 304: Based on the multi-factor authentication mechanism of the side chain, the verification process is analyzed and the ownership of the target authentication data is determined to generate a verification result.

[0050] As an optional embodiment, based on the multi-factor authentication mechanism of the side chain, the verification process is analyzed, and the ownership of the target authentication data is determined to generate a verification result, which may specifically include: Step 3041: decrypting the digital signature based on the creator's public key in the target authentication data and generating a signature validity identifier.

[0051] In this step, when generating the target authentication data, the creator uses his private key to sign the data. The signed data can be verified during the transmission process to ensure that the data has not been tampered with and is indeed from the creator; during the verification process, the creator's public key is used to decrypt the digital signature in the target authentication data. The public key and private key are a pair of keys. The public key is used for decryption and the private key is used for signing. By decrypting the signature, the validity of the signature can be verified; if the decrypted data is consistent with the target authentication data, it means that the signature is valid, and a signature validity identifier is generated.

[0052] Step 3042: Extract the biometric hash value from the target authentication data and perform similarity calculation with the registration features pre-stored in the off-chain security database to generate a biometric matching degree identifier.

[0053] In this step, the biometric hash value is a unique identifier generated by hashing the creator's biometric features (such as fingerprints, voiceprints, facial recognition, etc.), which is used to verify the creator's identity and ensure that the target authentication data matches the creator's biometric features; the biometric hash value is extracted from the target authentication data and the similarity is calculated with the registration features pre-stored in the off-chain security database, where the off-chain security database stores the creator's biometric hash value. If the biometric hash value in the target authentication data is highly similar to the registration features in the database, indicating that the biometric features match, a biometric matching degree identifier is generated, indicating that the creator's identity authentication is passed.

[0054] Step 3043: Access the external copyright registration system through the on-chain oracle module to query the off-chain ownership status based on the content fingerprint of the target authentication data and generate an ownership registration validity identifier.

[0055] In this step, the on-chain oracle module is a mechanism for the interaction between the blockchain and the external copyright registration system. It allows the blockchain smart contract to access external data sources, such as the copyright registration system. The external copyright registration system is accessed through the on-chain oracle module, and the content fingerprint in the target authentication data is used for query. The content fingerprint is the unique identifier of the target authentication data and is used to find the corresponding ownership information in the copyright registration system. The ownership registration validity identifier is generated based on the query result. If the copyright registration system confirms that the content fingerprint of the target authentication data has been successfully registered and the ownership is clear, a ownership registration validity identifier is generated to indicate that the ownership status of the target authentication data is valid.

[0056] Step 3044: If the signature validity flag, biometric matching flag and ownership registration validity flag all return to the passed status, a valid ownership verification result is generated and the evidence storage and chain operation is triggered; otherwise, an ownership exception alarm result containing a failure factor type is generated.

[0057] In this step, if the signature validity identifier, biometric matching identifier and ownership registration validity identifier all return to the passed status, it means that the target authentication data has been verified in terms of signature, biometrics and ownership registration, and the ownership validity verification result is generated at this time; after the ownership validity verification result is generated, the evidence storage and chain operation is triggered; if any one of the identifiers fails to pass the verification, an ownership exception alarm result containing a failure factor type is generated, where the failure factor type may include invalid signature, biometric mismatch or invalid ownership registration, etc. The ownership exception alarm result is used to prompt the administrator or relevant parties for further investigation and processing.

[0058] Step 305: If an ownership anomaly warning result is received, a derivative work identifier and a corresponding derivative work are generated based on the content fingerprint in the target authentication data.

[0059] In this step, if an abnormal ownership alarm result is received, indicating that the target authentication data has failed any verification link such as signature validity, biometric matching or ownership registration validity, the content fingerprint in the target authentication data will be further analyzed; the content fingerprint is the unique identifier of the target authentication data, which can uniquely identify the content of the target authentication data and is the basis for copyright review and traceability; a derivative work identifier is generated based on the content fingerprint to uniquely identify the derivative work related to the target authentication data. The derivative work refers to a new work formed by adaptation, translation, annotation, and compilation based on the original work.

[0060] Step 306: Store the derivative work identifier on the blockchain, and send the derivative work to the copyright detection node for secondary copyright review.

[0061] In this step, the generated derivative work identifier is stored on the blockchain as a basis for subsequent copyright review and traceability, providing support for resolving copyright disputes. The copyright detection node is a node specifically used for copyright review, responsible for conducting further copyright review of the target authentication data and its derivative works to confirm their ownership status. If the review passes, it indicates that the ownership of the derivative work is legal. If it fails, a new ownership anomaly alarm result will be generated, prompting further processing. This solution can not only verify the copyright ownership of the work to be authenticated, but also generate new derivative works for secondary authentication through adaptation of data that has not passed the verification. Adaptation and innovation are carried out under the premise of legality and compliance, which can stimulate the creativity of creators and promote the production of more excellent derivative works.

[0062] Step 40: Write the target authentication data that has passed the authentication and the verification result into the blockchain accordingly.

[0063] As an optional embodiment, the authenticated target authentication data and the verification result are written into the blockchain correspondingly, which may specifically include: Step 401: extracting the content fingerprint of the target authentication data, generating a random confusion factor and calculating the anonymized content hash.

[0064] In this step, the random obfuscation factor is a mechanism used to enhance privacy protection. By introducing randomness, the content fingerprint cannot be directly identified during storage and transmission, thereby protecting the privacy of the data. The content fingerprint is extracted from the target authentication data, and the content fingerprint is combined with the random obfuscation factor to generate an anonymized content hash through a hash function calculation. The anonymized content hash retains the uniqueness of the content fingerprint while enhancing privacy protection through the random obfuscation factor.

[0065] Step 402: Encrypt the target authentication data using a symmetric encryption algorithm and bind it with the anonymized content hash to generate a privacy evidence package.

[0066] In this step, the symmetric encryption algorithm is an encryption technology that uses the same key for encryption and decryption. It is efficient and secure and is suitable for encrypting large amounts of data. The encrypted target authentication data is bound to the anonymized content hash to generate a privacy evidence package, which contains the encrypted data and the anonymized content hash for subsequent storage and verification. Through encryption and anonymization processing, the privacy evidence package ensures the privacy and security of the target authentication data during storage and transmission.

[0067] Step 403: Write the privacy evidence package to the side chain through the cross-chain evidence network, and simultaneously write the anonymized content hash to the main chain.

[0068] In this step, the privacy evidence package is encapsulated according to the protocol requirements of the cross-chain evidence network through the cross-chain evidence network, and the encapsulated privacy evidence package is sent from the main chain to the side chain. After the side chain receives the privacy evidence package, it stores it in the distributed ledger of the side chain; when the privacy evidence package is written to the side chain, the anonymized content hash is written to the main chain. After the main chain receives the anonymized content hash, it stores it in the distributed ledger of the main chain. The anonymized content hash serves as the unique identifier of the target authentication data, ensuring the security and privacy of the data during storage and transmission, realizing distributed storage and trusted verification of data, and providing solid technical support for the copyright protection of digital content.

[0069] The above is an embodiment of the method proposed in this application. Based on the same inventive concept, this application embodiment also provides an AI digital content copyright notarization device based on blockchain, the structure of which is as follows: Figure 2 shown.

[0070] Figure 2 This is a schematic diagram of the internal structure of an AI digital content copyright notarization device based on blockchain provided in the embodiment of this application. Figure 2 As shown, the equipment includes: at least one processor 201; and, a memory 202 communicatively coupled to the at least one processor; Among them, the memory 202 stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor 201 to enable at least one processor 201 to: obtain data to be authenticated and store the data to be authenticated on the blockchain; wherein the data to be authenticated includes text data, image data, audio and video data; input the data to be authenticated into a preset multimodal feature extraction model to generate target authentication data; based on a preset smart contract, verify the ownership of the target authentication data and generate a verification result; write the authenticated target authentication data and the verification result into the blockchain accordingly.

[0071] Some embodiments of the present application provide corresponding Figure 1 A non-volatile computer storage medium for blockchain-based AI digital content copyright notarization stores computer executable instructions, wherein the computer executable instructions are configured to: obtain data to be authenticated and store the data to be authenticated on the blockchain; wherein the data to be authenticated includes text data, image data, and audio and video data; input the data to be authenticated into a preset multimodal feature extraction model to generate target authentication data; based on a preset smart contract, verify the ownership of the target authentication data and generate a verification result; and write the authenticated target authentication data and the verification result into the blockchain accordingly.

Claims

1. A blockchain-based AI digital content copyright notarization method, characterized by: The method comprises: Obtaining data to be authenticated and storing the data to be authenticated on a blockchain; wherein the data to be authenticated includes text data, image data, audio and video data; Inputting the data to be authenticated into a preset multimodal feature extraction model to generate target authentication data; Based on the preset smart contract, the ownership of the target authentication data is verified and a verification result is generated; The target authentication data that has passed the authentication is written into the blockchain corresponding to the verification result.

2. The blockchain-based AI digital content copyright notarization method according to claim 1 is characterized in that: Inputting the data to be authenticated into a preset multimodal feature extraction model to generate target authentication data, specifically including: Performing feature extraction on the data to be authenticated to obtain a multimodal feature vector; Based on the cross-modal attention mechanism, a gated fusion network is used to dynamically weight and aggregate the multimodal feature vectors to generate a joint feature vector; A hierarchical hash operation is performed on the joint feature vector to obtain target authentication data.

3. The blockchain-based AI digital content copyright notarization method according to claim 2 is characterized in that: Performing feature extraction on the data to be authenticated to obtain a multimodal feature vector, specifically including: Performing natural language feature encoding on the text data to generate a text feature vector; Performing visual feature encoding on the image data to generate an image feature vector; Performing spatiotemporal feature encoding on the audio and video data to generate an audio and video joint feature vector; The text feature vector, the image feature vector and the audio and video joint feature vector are correlated to obtain the multimodal feature vector.

4. The blockchain-based AI digital content copyright notarization method according to claim 2 is characterized in that: Based on the cross-modal attention mechanism, the multimodal feature vectors are dynamically weighted and aggregated using a gated fusion network to generate a joint feature vector, specifically including: Based on the cross-modal attention mechanism, a cross-modal attention score between the text feature vector, the image feature vector, and the audio and video joint feature vector is calculated, and a weight coefficient corresponding to the cross-modal attention score is generated; Inputting the weight coefficient into a gated fusion network, and dynamically modifying the weight coefficient based on a preset gating mechanism in the gated fusion network; Based on the modified weight coefficients, weighted summation is performed on the multimodal feature vectors to generate the joint feature vector.

5. The blockchain-based AI digital content copyright notarization method according to claim 1 is characterized in that: Based on the preset smart contract, the target authentication data is verified for ownership and a verification result is generated, specifically including: Based on the preset main chain-side chain cross-chain evidence network transmission protocol, the target authentication data is sent from the side chain to the main chain; According to the verification consensus rules of the main chain, basic verification is performed on the format integrity and data source legitimacy of the target authentication data; After the basic verification is passed, the smart contract preset in the main chain is mobilized to send the verification process data back to the side chain through the cross-chain evidence storage network transmission protocol; Based on the multi-factor authentication mechanism of the side chain, the verification process is analyzed, and the ownership of the target authentication data is determined to generate a verification result.

6. The blockchain-based AI digital content copyright notarization method according to claim 5 is characterized in that: Based on the multi-factor authentication mechanism of the side chain, the verification process is analyzed and the ownership of the target authentication data is determined to generate a verification result, specifically including: Decrypting the digital signature based on the creator's public key in the target authentication data to generate a signature validity identifier; Extract the biometric hash value from the target authentication data and perform similarity calculation with the registered features pre-stored in the off-chain security database to generate a biometric matching indicator; Access the external copyright registration system through the on-chain oracle module to query the off-chain ownership status based on the content fingerprint of the target authentication data and generate an ownership registration validity identifier; If the signature validity identifier, the biometric matching identifier and the ownership registration validity identifier all return a passed status, a valid ownership verification result is generated and the evidence storage and chain operation is triggered; otherwise, an ownership abnormality alarm result containing a failure factor type is generated.

7. The blockchain-based AI digital content copyright notarization method according to claim 1 is characterized in that: After generating the verification result, the method further includes: If the ownership anomaly warning result is received, generating a derivative work identifier and a corresponding derivative work based on the content fingerprint in the target authentication data; The derivative work identifier is stored on the blockchain, and the derivative work is sent to the copyright detection node for secondary copyright review.

8. The blockchain-based AI digital content copyright notarization method according to claim 1 is characterized in that: Writing the authenticated target authentication data and the verification result into the blockchain in correspondence, specifically including: Extracting a content fingerprint of the target authentication data, generating a random confusion factor and calculating an anonymized content hash; Encrypt the target authentication data using a symmetric encryption algorithm and bind it with the anonymized content hash to generate a privacy evidence package; The privacy evidence package is written into the side chain through the cross-chain evidence network, and the anonymized content hash is simultaneously written into the main chain.

9. An AI digital content copyright notarization device based on blockchain, characterized in that: The device comprises: at least one processor; and, a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform a method according to any one of claims 1 to 8.

10. A non-volatile computer storage medium for AI digital content copyright notarization based on blockchain, storing computer-executable instructions, characterized in that: When the computer-executable instructions are executed, a method according to any one of claims 1 to 8 is implemented.

Images