Dynamic key threshold fragmentation privacy tracing method based on blockchain collaborative verification

Abstract

The application discloses a dynamic key threshold fragmentation privacy tracing method based on blockchain collaborative verification and belongs to the field of information security. The method takes the dynamic key as the pivot, deeply couples the decentralized identity, zero-knowledge verification, deep watermark and threshold collaborative decryption, splits the temporary key generated each time into multiple fragments, and stores the fragments in a distributed manner by the blockchain node and the tracing unit. Only the hash fingerprints of the combined fragments are stored on the chain to form a lightweight evidence chain which is tamper-proof but does not expose privacy. Participants access anonymously through verifiable credentials without leaking specific information, and the authenticity of the anonymous identity is ensured by the signature of the trusted unit. When data leakage occurs, the tracing unit extracts the watermark, compares the zero-knowledge proof, verifies the relevance of the leaked ciphertext and the on-chain fingerprint, triggers the multi-party threshold decryption to recover the key, and finally accurately locates the person responsible by combining the encrypted stored DID mapping table, without exposing irrelevant information throughout the process.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information security, and in particular to a dynamic key threshold fragmentation privacy tracing method based on blockchain collaborative verification. BACKGROUND

[0002] With video conferencing becoming the core scenario of digital collaboration, the dual needs of data privacy and trusted evidence have become increasingly prominent. Traditional centralized identity management systems rely on a single trusted unit to centrally manage and control user identity information. This mode not only has the risk of single point of failure, but also is difficult to support cross-domain interoperability requirements. Although the decentralized identity (DID) system catalyzed by blockchain technology gives users the right to manage their own identity through public-private key pairs, existing solutions still face a dilemma between privacy protection and compliance auditing. Completely anonymous DID identifiers are difficult to meet regulatory tracing requirements, while directly mapping identity relationships in plaintext on-chain will expose sensitive information. This contradiction is particularly prominent in the video conference scenario: traditional solutions rely on centralized servers to encrypt audio and video streams and audit logs, but service providers can tamper with conference records, and password-based identity authentication mechanisms are also vulnerable to man-in-the-middle attacks, leading to ineffective responsibility tracing after content leakage.

[0003] Existing blockchain and decentralized identity solutions face deep technical contradictions between privacy protection and regulatory tracing. The current mainstream implementation approach generally faces a dilemma: if identity association information is stored in its entirety on-chain to ensure traceability, the public key-identity mapping relationship will be exposed in the public network, which seriously violates the principles of data minimization and privacy protection. While a completely anonymous DID system can protect user identity privacy, it completely loses the ability to trace responsibility when unauthorized screen capture and other leakage events occur in video conferences. More complex is the end-to-end encryption requirement specific to the video conference scenario, which makes the traditional auditing mechanism face a dilemma - if the service provider cannot obtain the session key, it is difficult to track the leakage path, and using a centralized key escrow solution will introduce a single point of trust risk. These deep-seated technical contradictions make it difficult for existing solutions to meet the basic requirements of judicial evidence and the compliance standards of data protection regulations, and there is an urgent need to build a new technical architecture that can achieve effective tracing while protecting privacy.

[0004] Liu et al. (Liu Z, Yu X, Liu N, et al. Integrating AI with detection methods, IoT, and blockchain to achieve food authenticity and traceability from farm-to-table [J]. Trends in Food Science & Technology, 2025: 104925.) proposed an innovative approach that combines artificial intelligence (AI), the Internet of Things (IoT), and blockchain technology to optimize the entire process of data collection and management. In this solution, AI technology is used for intelligent analysis and decision support, IoT technology enables extensive physical device connection and data collection, and blockchain provides an unalterable and traceable guarantee for data storage and transmission. Through this integrated architecture, not only the efficiency and accuracy of data collection are improved, but also the security and credibility of traceable information are significantly enhanced.

[0005] Brandín et al. (Brandín R, Abrishami S. IoT-BIM and blockchain integration for enhanced data traceability in offsite manufacturing [J]. Automation in Construction, 2024, 159: 105266.) proposed a Supply Chain Management Model based on information modeling (SCMM-PDO) and built a highly secure and transparent traceability management framework. This model standardizes and uniformly manages data from different supply chain nodes through information modeling technology, and realizes data credible sharing and traceability among multiple parties with the help of blockchain's unalterable and transparent nature. At the same time, SCMM-PDO designs a perfect data access and authorization mechanism to ensure the protection of privacy and commercial secrets of all parties under the premise of data openness and transparency.

[0006] Although the above research has made significant progress in improving data transparency and traceability, most of the schemes mainly focus on the authenticity and integrity of on-chain data, and pay insufficient attention to the protection of data privacy. Due to the openness of the blockchain, the public storage and transmission of sensitive data on the chain may lead to the risk of privacy leakage. Therefore, how to protect user privacy while ensuring traceability has become a problem to be solved in the current blockchain traceability forensics technology. SUMMARY

[0007] The purpose of the application is to provide a dynamic key threshold fragmentation privacy tracing method based on blockchain collaborative verification.

[0008] Technical scheme: The dynamic key threshold fragmentation privacy tracing method based on blockchain collaborative verification of the application comprises the following steps:

[0009] The participating user registers for identity, generates a master key pair and a DID identifier;

[0010] The trusted unit issues a verifiable certificate for the participating user according to the DID identifier;

[0011] The participating user generates zero-knowledge proof verification information when participating in the meeting, and the system verifies the identity of the participating user according to the generated zero-knowledge proof verification information;

[0012] Use Unet for encoding on the conference system, and use a symmetric decoder with a shared encoder downsampling parameter for decoding;

[0013] Store the key information to the blockchain;

[0014] When information leakage occurs, trigger the traceability and identity tracking mechanism, search and compare on the blockchain, and decrypt the tracked identity.

[0015] Further, the step of the participating user registering for identity and generating a master key pair and a DID identifier comprises:

[0016] The participating user generates a master key pair locally before participating in the meeting for the first time ), wherein is a master private key, is a master public key;

[0017] Generate a master virtual identity information = according to the master public key, wherein represents a hash function, represents a timestamp, represents a random number, represents a link operation;

[0018] Generate a key through threshold distributed key and adopt Threshold encryption algorithms use keys Disassembled Each segment It requires at least It takes collaboration among multiple fragments to recover the key, among which Indicates the first One fragment;

[0019] Constructing encrypted tuples for secure storage And stored on the authentication chain, where the encrypted tuple is represented as , It is the AES threshold encryption function. This indicates the real identity of the participating users;

[0020] When users who are not first-time participants join a meeting, they can generate a sub-public key pair in a secure or offline environment based on the current meeting information or time and the master key pair. ,in For the child's private key, The child public key is represented as follows: , In the formula, Indicates time parameter, This represents the modulo operation, where n represents the order of the elliptic curve. This represents the public key generation function. Represents the generator of an elliptic curve;

[0021] Participants randomly generate salt values ​​through the conference system. Encrypt the salt value to obtain the encrypted salt value. ;

[0022] Calculate intermediate binding parameters The formula is: ;

[0023] The virtual identity information for this meeting is generated by combining the sub-public key, session context, and intermediate hash value. ,in Indicates the context.

[0024] Furthermore, the expression for the verifiable credential is:

[0025] ,

[0026] In the formula, It is an attribute strategy, with an input context. Using functions Determine whether dynamic generation is allowed. , Indicates the range of two variables; Indicate the start and end times; It is a BLS signature generated by a trusted unit. It is the private key of the trusted unit. It is a multiplication cyclic group.

[0027] Furthermore, during the meeting, participating users generate zero-knowledge proof verification information. The system then verifies the identity of the participating users based on this generated zero-knowledge proof verification information, including the following steps:

[0028] Build zero-knowledge proof circuits locally on the user's device The inputs, including public inputs and private inputs, are represented as follows:

[0029] ,

[0030] in, express The hash value;

[0031] Design constraints include:

[0032] ,

[0033] ,

[0034] ,

[0035] in, The public key representing a trusted unit, express generator, Represents a bilinear mapping , It is a multiplication cyclic group. It is the target multiplication group;

[0036] The final zero-knowledge proof verification information is constructed as follows:

[0037] ,

[0038] In the formula, This represents a function that generates a cryptographic proof of π.

[0039] The system verifies the triple legitimacy of participating users' identities, including:

[0040] Identity ownership: Users do indeed own the generated dynamics Master private key ;

[0041] dynamic Legitimacy: Dynamic identity is based on , 、 、 correctly generated;

[0042] Identity authentication validity: the authentication signature held by the user is legal; the verification equation is wherein represents a verification function.

[0043] Further, using Unet to encode on the conference system, the step of decoding using a symmetric decoder that shares the down-sampling parameters of the encoder includes:

[0044] Step 41, in the message processor, input the secret message , transform the secret message into a three-dimensional tensor , denoted as: wherein, represents a dimension reshaping operation, represents that the tensor element takes a value in a real number field, represents the length of the secret message, represents the number of message channels, feature map height and width;

[0045] Then use convolution kernel to improve feature expression capability, get feature F1, denoted as:

[0046] wherein, represents a convolution, batch normalization and activation function atomic operation;

[0047] Then use a deconvolution layer with a step of 2 to gradually expand the resolution to the target size , and optimize the features through SE modules , get feature , the formula is:

[0048] ,

[0049] wherein, respectively represent the number of channels, height and width of the carrier image;

[0050] Finally, through the convolution layer, the number of channels is aligned with the carrier image, and the processed secret message is obtained, denoted as:

[0051] , wherein, represents a convolution layer;

[0052] Step 42, in the improved encoder operation, input the carrier image , and the secret message after message processing Channel concatenation is performed to obtain fused information, denoted as:

[0053] wherein, denotes the channel dimension;

[0054] The fused information is input into the Unet architecture, and then stepwise down-sampling is performed to obtain information , denoted as:

[0055] ,

[0056] wherein, denotes the i-th layer stepwise down-sampling module with a stride of 2, denotes that the encoder ends at the i-th layer;

[0057] Then, the image size is stepwise restored through an up-sampling module, and the encoded features are incorporated into the current feature map using a skip connection to obtain a feature map , and the formula is:

[0058] ,

[0059] wherein, denotes a deconvolution layer with a stride of 2, denotes the i-th convolution function;

[0060] The final output is an image embedded with a watermark , and the formula is:

[0061] ;

[0062] Step 43, noise layer: a small batch training strategy Mini-Batch is adopted, and each time in the training, a random selection is made in the real non-differentiable noise attack layer, the noise-free attack layer and the non-differentiable simulated noise layer, denoted as:

[0063] ,

[0064] wherein, denotes the noise layer;

[0065] Step 44, decoder: a symmetric encoding structure is adopted, and a symmetric decoder sharing the down-sampling parameters of the encoder is used to extract the watermark information, and the shared parameter layer, wherein denotes the input feature map of the i-th layer in the decoder down-sampling path, and the formula is:

[0066] ,

[0067] wherein, represents the input feature map of the layer;

[0068] Spatial aggregation and output:

[0069] ,

[0070] ,

[0071] wherein, represents a 1x1 convolution layer, represents a single-channel feature map after convolution, represents the extracted secret message, represents a Sigmoid function.

[0072] Further, when information leakage occurs, a traceability and identity tracking mechanism is triggered, and the steps of searching and comparing on the blockchain to decrypt and track the identity include:

[0073] Step 61, extract the watermark information in the leaked content, compare it with the stored watermark information on the chain, lock the specific meeting, and the virtual identity information generated by the participant in this meeting ;

[0074] Step 62, through the preset zero-knowledge verification and Merkle tree verification, prove that the is valid and participates in the current meeting;

[0075] Step 63, aggregate and decrypt the key , decrypt , get value, after searching and comparing on the chain, decrypt to track the identity ID: , wherein represents a decryption salt value operation, represents an intermediate binding parameter in decryption.

[0076] Further, the key information includes meeting ID, screen watermark hash value, encrypted salt value, intermediate binding parameter, hash value of verifiable certificate, zero-knowledge proof, participant's temporary virtual identity information and constructed Merkle tree root value, off-chain information storage address and timestamp information.

[0077] Beneficial effects: compared with the prior art, the present application has the following advantages:

[0078] 1. The application realizes end-to-end encryption protection and verifiable traceability mechanism of video conference full-process data through the collaborative architecture of blockchain storage and dynamic key threshold fragmentation, and guarantees the user privacy sovereignty.

[0079] 2. To prevent the problem that the responsible subject cannot be tracked after the conference content is leaked, the application innovatively designs a dynamic identity binding strategy, a verifiable certificate system based on zero-knowledge proof, and a coding and decoding fusion watermark based on Unet architecture: participants access the meeting through dynamically generated anonymous identities , whose real identities are encrypted and anchored off-chain by verifiable certificates (VC) issued by trusted units; only the hash fingerprints of session key fragments and zero-knowledge proof evidence exist on-chain, ensuring that attackers cannot reverse sensitive information through on-chain data;

[0080] 3. When data leakage occurs, the tracing unit determines the identity information of the participants through the preset zero-knowledge verification circuit, and through on-chain and off-chain information comparison, triggers multi-party collaborative threshold decryption to recover the key, and accurately locates the responsible person based on the encrypted identity mapping table.

[0081] 4. The application realizes multiple protections in the aspects of anonymous access, data minimization on-chain, key fragment disaster recovery, etc. through deep coupling of cryptographic primitives, which not only eliminates the risk of single-point tampering and single-key leakage in centralized systems, but also avoids the traceability failure caused by complete anonymity, finally builds a governance paradigm of privacy available and invisible, traceable and controllable, and provides a safe and compliant dynamic balance base for high-sensitivity digital collaboration scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0082] Figure 1 is a user registration on-chain flowchart;

[0083] Figure 2 is a temporary identity generation schematic diagram;

[0084] Figure 3 is a coding and decoding fusion watermark based on Unet architecture;

[0085] Figure 4 is a traceability time consumption diagram of each step. DETAILED DESCRIPTION

[0086] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application is further described in detail below in combination with the drawings and examples.

[0087] The application aims to build a video conference security architecture balancing privacy protection and compliance traceability, and realizes a trust mechanism of "data available but invisible" through cryptographic technology. The core innovation is to deeply couple decentralized identity, zero-knowledge proof, deep watermark and threshold collaborative decryption with dynamic key as the pivot: the temporary key generated each time is split into multiple fragments, which are distributed by blockchain nodes and traceability units, only the hash fingerprint of the combined fragments is stored on the chain to form a lightweight evidence chain that is tamper-proof but does not expose privacy. Participants access anonymously through verifiable credentials, such as proving themselves to be "legitimate participants" without revealing specific information, and the authenticity of anonymous identity is ensured by trusted unit signature. When data leakage occurs, the traceability unit extracts the watermark, compares the zero-knowledge proof, verifies the relevance of the leaked ciphertext and the on-chain fingerprint, triggers multi-party threshold decryption to recover the key, and finally locates the responsible person accurately combined with the encrypted DID mapping table, without exposing irrelevant information throughout the process. This architecture keeps audio and video streams and metadata in end-to-end encryption state in daily meetings, and the traceability process replaces traditional manual audit with mathematical verifiability, meeting the "right to be forgotten" requirement of GDPR (General Data Protection Regulation), and quickly extracting evidence when disputes occur, providing a dynamic balance between privacy and compliance for multi-party collaboration.

[0088] Through the above technical innovation, the application creatively deeply couples decentralized identity, zero-knowledge proof, deep watermark and threshold cryptography, builds a three-layer governance architecture (identity management layer, data evidence layer, security traceability layer) with privacy priority and controllable exceptions, and solves the natural contradiction between privacy protection and regulatory traceability in traditional solutions. The trust paradigm of digital collaboration is restructured, and the fragile trust chain relying on institutional endorsement in the centralized era is transformed into a decentralized and verifiable trust network guaranteed by mathematical algorithms. All data flows in daily meetings exist in ciphertext form, and service providers and even internal personnel cannot break through the cryptographic barrier to obtain sensitive information; once a security incident occurs, the preset cryptographic rules automatically activate the traceability process to extract legal evidence on the premise of protecting the privacy of irrelevant parties. This "privacy-first, exception-controlled" governance model not only solves the long-standing security and efficiency paradox in video conferencing, but also provides a reusable privacy and security framework for emerging scenarios such as industrial Internet of Things, promoting digital collaboration to a higher level of trust dimension.

[0089] The dynamic key threshold fragmentation privacy traceability method based on blockchain collaborative verification described in the embodiment includes the following steps:

[0090] Step 1, the participant user registers for identity, generates a master key pair and a DID identifier.

[0091] In combination Figure 1, initially need to conduct conference creation and identity registration, further, the step of identity registration of the participating user, generating the master key pair and DID identifier includes:

[0092] The participating user generates a master key pair in a local trusted execution environment TEE before participating in the conference for the first time ), wherein is a master private key, is a master public key;

[0093] A master virtual identity information is generated according to the master public key = , wherein, denotes a hash function, denotes a timestamp, denotes a random number, denotes a link operation;

[0094] A key is generated by threshold distributed key generation , and the key is divided into pieces by using a threshold encryption algorithm , and at least pieces are required to recover the key, wherein denotes the piece, and the number m depends on the number of traceable units and related personnel;

[0095] An encrypted tuple for secure storage is constructed and stored on the identity verification chain, wherein the encrypted tuple is represented as , is an AES (Advanced Encryption Standard) threshold encryption function, denotes the real identity of the participating user, and the encrypted key is distributed to each traceable unit and related person in charge;

[0096] When a non-first-time participating user participates in the conference, the user can generate a sub-public key pair in a secure environment or offline environment according to the current conference information or time by means of the master key pair, wherein is a sub-private key, is a sub-public key, and is respectively represented as: , , denotes a time parameter, denotes a modulo operation, n denotes the order of an elliptic curve, denotes a sub-public key generation function, denotes an elliptic curve generator;

[0097] The participant generates a random salt value through the conference system The salt value is encrypted to obtain an encrypted salt value ;

[0098] To bind and the salt value, an intermediate binding parameter is calculated to prevent malicious link behavior, the formula is: ;

[0099] The virtual identity information of the conference is generated by combining the child public key, the session context and the intermediate hash value , wherein represents the context environment.

[0100] Step 2, the trusted unit issues a verifiable credential for the participant according to the DID identifier.

[0101] The above trusted unit refers to a trusted entity, such as a system for providing verification and authentication services.

[0102] In the credential distribution phase, the trusted unit issues a verifiable credential for the participant .

[0103] Further, the expression of the verifiable credential is:

[0104] ,

[0105] In the formula, is the attribute policy, the input context environment , and the policy definition function is used to determine whether to allow dynamic , wherein , P, S, and T represent user identity permissions, target scenarios, and time windows, respectively, represents a binary value domain; represents the start time and the end time; is a BLS signature generated by the trusted unit, is the private key of the trusted unit, is a multiplicative cyclic group.

[0106] As shown in Figure 2 , in the conference participation key fragmentation phase, the system will generate a key for the conference through threshold distributed key generation , and use threshold encryption algorithm to divide it into n fragments , and at least fragments are needed to recover the key, wherein represents the nth slice, the number of n depends on the number of tracing units and related personnel. The user will create a random salt value , and use threshold encryption: . In order to bind And the salt value, it is necessary to calculate the intermediate binding parameter: . Finally generate the of this meeting.

[0107] Step 3, when attending, the attending user generates zero-knowledge proof verification information, and the system verifies the identity of the attending user according to the generated zero-knowledge proof verification information.

[0108] Further, when attending, the attending user generates zero-knowledge proof verification information, and the system verifies the identity of the attending user according to the generated zero-knowledge proof verification information. The steps include:

[0109] Construct the input of the zero-knowledge proof circuit locally by the user, including public input and private input, respectively represented as:

[0110] ,

[0111] Among them, represents the hash value of ;

[0112] The design constraint condition is:

[0113] ,

[0114] ,

[0115] ,

[0116] Among them, represents the public key of the trusted unit, represents the generator of , represents the bilinear mapping , is a multiplicative cyclic group, is a target multiplicative group;

[0117] Finally, the zero-knowledge proof verification information is constructed, represented as:

[0118] ,

[0119] In the formula, represents a function of generating a cryptographic proof π;

[0120] The system verifies the triple legality of the identity of the attending user, including:

[0121] Identity ownership: the user indeed owns the dynamic identity generated by the master private key ;

[0122] Dynamic identity legitimacy: the dynamic identity is based on , , , correct generation

[0123] Identity authentication validity: the authentication signature held by the user is legal; the verification equation is , wherein represents a verification function.

[0124] Through this method, it can be verified that the dynamic DID generated by the user is legal without exposing specific information, and the identity attribute behind it is authenticated by a trusted unit.

[0125] Step 4: Encode using Unet on the conference system, and decode using a symmetric decoder sharing the down-sampling parameters of the encoder.

[0126] In the encoding-decoding fusion watermarking stage based on the Unet architecture, the existing watermarking architecture mainly includes an encoder, an information processor, a noise layer, a decoder, and a discriminator. In order to make the screen watermark have higher robustness and invisibility, make it more efficient and stable to extract information, and enhance the stability of subsequent tracking, the original encoding-decoding separation architecture is redesigned in this example. Unet is used for encoding, and then decoding uses a symmetric decoder sharing the down-sampling parameters of the encoder, sharing parameters, and enhancing the coupling ability of encoding and decoding.

[0127] As shown in Figure 3 , further, the step of encoding using Unet on the conference system and decoding using a symmetric decoder sharing the down-sampling parameters of the encoder includes:

[0128] Step 41: In the message processor, input the secret message , the secret message including conference information and user virtual identity information in this conference, transform the secret message into a three-dimensional tensor , represented as: , wherein represents a dimension reshaping operation, represents that the tensor element takes a value in a real number field, represents the length of the secret message, represents the number of message channels, feature map height, and width;

[0129] Again use Convolutional kernels enhance feature representation capabilities, yielding the F1 feature, denoted as:

[0130] ,in, Represents atomic operations for convolution, batch normalization, and activation functions;

[0131] Next, deconvolutional layers with a stride of 2 are used to gradually expand the resolution to the target size. ,like and through SE module Optimize features to obtain features The formula is:

[0132] ,

[0133] In the formula, These represent the number of channels, height, and width of the carrier image, respectively.

[0134] Finally, a convolutional layer is used to align the number of channels with the carrier image to obtain the processed secret message. , is represented as:

[0135] ,in, This indicates a convolutional layer.

[0136] Step 42, in the improved encoder operation, input carrier image ,Will and secret messages after message processing By performing channel splicing, the merged information is obtained, represented as follows:

[0137] ,in, Indicates the channel dimension;

[0138] The merged information The information is input into the Unet architecture and then progressively downsampled to obtain the data. , is represented as:

[0139] ,

[0140] in, Indicates the first A convolutional downsampling module with a stride of 2. This indicates that the encoder ends at the th . layer;

[0141] Next, the image size is gradually restored through the upsampling module, and skip connections are used to incorporate the encoded features into the current feature map to obtain the feature map. The formula is:

[0142] ,

[0143] wherein, denotes a deconvolution layer with step size 2, denotes the step convolution function;

[0144] the final output is the image with embedded watermark is:

[0145] .

[0146] Step 43, the noise layer solves the training-test inconsistency problem in the deep watermarking model, and improves the robustness to JPEG compression. The traditional method uses a fixed noise layer, which leads to model overfitting to a specific attack mode, while through dynamic mixing of noise sources, the model is forced to learn the general anti-attack ability. In this example, the Mini-Batch strategy is used for small batch training. At each training time, a random selection is made among the real non-differentiable noise attack layer, the noise-free attack layer and the non-differentiable simulated noise layer, denoted as:

[0147] ,

[0148] wherein, denotes the noise layer.

[0149] Step 44, decoder: responsible for reconstructing the secret message from the noisy image , adopts a symmetric encoding structure, uses the down-sampling process in the encoder to extract watermark information, reuses the down-sampling part of the encoder, and shares the parameter layer, wherein denotes the input feature map of the layer in the decoder down-sampling path, denoted as:

[0150] ,

[0151] wherein, denotes the input feature map of the layer;

[0152] Spatial aggregation and output:

[0153] ,

[0154] ,

[0155] wherein, denotes the single-channel feature map after convolution, denotes the extracted secret message, denotes the Sigmoid function.

[0156] Step 5, store the key information to the blockchain.

[0157] Finally, store the key information on the chain, including meeting ID, screen watermark hash value, encrypted salt value , intermediate binding parameters , VC hash value, zero-knowledge proof, participant temporary and constructed merkle tree root value, off-chain information storage address and timestamp information.

[0158] And off-chain, using IPFS (InterPlanetary File System, InterPlanetary File System) to store detailed information.

[0159] In the conference data encryption and evidence storage stage, the relevant information of the conference such as video, audio, text and embedded watermark information will be encrypted by threshold encryption algorithm to generate ciphertext , where is the dynamic key of this meeting, and message represents the relevant conference information. Finally, the on-chain evidence content adopts a lightweight structure:

[0160] ,

[0161] Among them, represents the unique identifier of the meeting, represents the screen camera watermark hash value, represents the root node of the merkle tree generated by all participants , represents the IPFS address stored off-chain, and T represents the timestamp.

[0162] Step 6, when information leakage occurs, trigger the traceability and identity tracking mechanism, search and compare on the identity verification chain, and decrypt the identity.

[0163] Further, when information leakage occurs, trigger the traceability and identity tracking mechanism, search and compare on the identity verification chain, and decrypt the identity. The steps include:

[0164] Step 61, first extract the watermark information in the leaked content and compare it with the watermark information stored on the identity verification chain to lock the specific meeting and the virtual identity information generated by the participant in this meeting ;

[0165] Step 62, through the preset zero-knowledge verification and merkle tree verification, prove that the locked is valid and participated in the current meeting;

[0166] Step 63, through multi-party cooperation aggregation decryption key , decrypting out , get value, after comparison on chain, decrypting tracking identity ID: , wherein represents a decryption salt value operation, represents a decryption intermediate binding value operation.

[0167] In one example, in the leakage time tracing phase, when the meeting content leakage is detected, the tracing unit will perform the following procedures to track the identity:

[0168] (a) Extract the screen watermark hash from the leaked data Match with on-chain , determine the associated meeting record and the corresponding suspicious user;

[0169] (b) Call the pre-prepared zero-knowledge verification circuit to verify the spatiotemporal consistency of extracted from the leaked ciphertext watermark in the on-chain Merkle tree, to ensure that the extracted is truly existent and complies with the regulations, and that the user did indeed participate in this meeting;

[0170] (c) Aggregate at least key fragments to perform threshold decryption, and through the decryption intermediate binding value , trace back to the user's on-chain real identity mapping table, thereby tracking to the specific person.

[0171] To further illustrate the effectiveness and superiority of the dynamic key threshold fragmentation privacy tracing method based on blockchain collaborative verification described in the present application, the time consumption of each step is experimentally determined. The experimental environment uses an i5-13490F processor and an NVIDIA GeForce RTX 4060ti GPU hardware configuration, runs an Ethereum blockchain network under an Ubuntu 20.04 LTS system, and completes cryptographic operation verification based on OpenSSL 3.0.8 and Groth16 protocol ZKP circuit. All experiments are conducted under the conditions of controlling network delay (0.1-0.5ms) and enabling GPU acceleration. The experimental results are shown in Figure 4 , in the screen watermark hash matching, image watermark extraction and on-chain hash matching consume 57ms and 15ms respectively; in the threshold decryption identity tracing, the fragmented aggregation request, threshold decryption calculation and on-chain identity mapping consume 73ms, 102ms and 17ms respectively; in the spatiotemporal consistency zero-knowledge verification, the virtual identity retrieval and ZK circuit verification consume 52ms and 154ms respectively. ​

[0172] In this embodiment, the gas consumption of the smart contract of traceability forensics is calculated, as shown in Table 1, and it can be known that the application consumes low on the chain. And the security comparison with the existing blockchain traceability method is shown in Table 2, and it can be known that the application has obvious advantages in security compared with the existing scheme.

[0173] Table 1

[0174]

[0175] Table 2

[0176]

[0177] In the application, the coding-decoding fusion watermark method based on the Unet architecture has significant advantages in robustness and concealment. In order to simulate the loss of watermark information caused by the change of shooting angle or framing in screen shooting (i.e. crop attack), the performance under three crop ratios of 30%, 50% and 70% is tested, and the test results are shown in Table 3. Although the cropping seriously destroys the integrity of the image, the application maintains a low bit error rate (BER) under all ratios, and maintains a high level of peak signal-to-noise ratio (PSNR) and structural similarity (SSIM), which proves that even in the case of partial information loss, the watermark content can be effectively recovered, and the outstanding screen camera resistance is exhibited. In comparison with HiDDeN, MBRS, Adaptor and other methods, the application has significant advantages in this test.

[0178] Table 3

[0179] .

Claims

1. A dynamic key threshold sharding privacy tracing method based on blockchain collaborative verification, characterized in that, The method comprises the following steps: The conference participant registers and generates a master key pair and a DID identifier; The trusted unit issues a verifiable credential to the conference participant according to the DID identifier; The conference participant generates zero-knowledge proof verification information when attending the conference, and the system verifies the identity of the conference participant according to the generated zero-knowledge proof verification information; Unet is used for encoding on the conference system, and a symmetric decoder with down-sampling parameters of a shared encoder is used for decoding; Key information is stored in a blockchain; When information leakage occurs, a traceability and identity tracking mechanism is triggered, and the blockchain is searched and compared to decrypt and track the identity; The step of using Unet for encoding on the conference system and using a symmetric decoder with down-sampling parameters of a shared encoder for decoding comprises: Step 41, in the message processor, input the secret message , the secret message is transformed into a three-dimensional tensor , denoted as: , where, denotes a dimension reshaping operation, denotes that the tensor element takes values in the real number field, denotes the length of the secret message, denotes the number of message channels, feature map height and width; Reusing The convolution kernel promotes the feature expression ability, and obtains a feature F1, which is expressed as: wherein, denote atomic operations of convolution, batch normalization, and activation function; Then a deconvolution layer with a step of 2 is used to gradually expand the resolution to the target size and through one SE module optimizing the features, the features , the formula is: ， In the formula, respectively represent the number of channels, height and width of the carrier image. Finally, the channel number is aligned with the carrier image by a convolution layer, and the processed secret message is obtained is represented as: wherein denotes a convolutional layer; Step 42, in the improved encoder operation, input carrier image , and the secret message after message processing and the secret message after message processing Channel splicing is performed to obtain the fused information, denoted as: wherein, denotes the channel dimension; The fused information is input into the Unet architecture, which is then progressively downsampled to obtain information is represented as: ， wherein, represents the end of the encoder at the layer convolutional down-sampling module with a stride of 2, represents the end of the encoder at the layer; Then the image size is gradually restored by the upsampling module, and the encoded features are incorporated into the current feature map using the skip connection to obtain the feature map , the formula is: ， wherein, denotes a deconvolution layer with a step size of 2, denotes the step convolution function; Final output the image with the watermark embedded To: ； Step 43, noise layer: a small batch training strategy Mini-Batch is adopted, and each time in the training, a random selection is made in the real non-differentiable noise attack layer, the noiseless attack layer and the non-differentiable simulated noise layer, denoted as: ， wherein represents a noise layer; Step 44, Decoder: using a symmetric decoding structure, extract watermark information using a symmetric decoder sharing the encoder down-sampling parameters, shared parameter layer, wherein represents the input feature map of the i-th layer in the decoder down-sampling path, is represented as: represents the input feature map of the i-th layer in the decoder down-sampling path, is represented as: ， wherein, represents the input feature map of the layer; Spatial aggregation and output: ， ， wherein, denotes a 1x1 convolutional layer, denotes a single-channel feature map after convolution, denotes the extracted secret message, denotes a sigmoid function; When information leakage occurs, a traceability and identity tracking mechanism is triggered, and the blockchain is searched and compared to decrypt and track the identity; Step 61, extract the watermark information in the leaked content, compare with the stored watermark information on the chain, lock the specific meeting, and the virtual identity information generated by the participants in this meeting ; Step 62, through the preset zero-knowledge proof and Merkle tree verification, prove that the locked is valid and participates in the current meeting; Step 63, aggregate decryption key by multi-party collaboration , decrypt 、 get value, search and compare on chain, decrypt trace identity ID: , where represents decryption salt value operation, represents decryption intermediate binding parameters.

2. The method of claim 1, wherein, The step of the conference participant registering and generating a master key pair and a DID identifier comprises: The attending user generates a master key pair locally before attending the conference for the first time wherein is a master private key, is a master public key; Generating a master virtual identity information from a master public key = wherein, denotes a hash function, denotes a time stamp, denotes a random number, denotes a concatenation operation; Key generation by threshold distributed key , and using A threshold encryption algorithm splits the key into shards , and at least shards are needed to recover the key, where denotes the th shard; Constructing a secure storage encrypted tuple and stored on the chain of identity authentication, wherein the encrypted tuple is represented as , is an AES threshold encryption function, represents the real identity of the participating user; When a non-first-time participant user participates in a conference, according to current conference information or time, a sub-public key pair is generated in a secure environment or an offline environment by means of a master key wherein is a sub-private key, is a sub-public key, and are respectively represented as: , , wherein represents a time parameter, represents a modulo operation, and n represents an order of an elliptic curve, represents a sub-public key generation function, represents an elliptic curve generator A participant generates a random salt value through a conference system The salt value is encrypted to obtain an encrypted salt value ; Computing intermediate binding parameters , the formula is: ; A virtual identity for the meeting is generated here from the joiner public key, the session context and the intermediate hash value wherein denotes the context environment.

3. The method of claim 2, wherein, The expression of the verifiable credential is: ， In the formula, is an attribute policy, an input context environment , using a function to determine whether to allow dynamic generation , represents a binary value domain; represents the start time and end time; is a BLS signature generated by a trusted unit, is a private key of a trusted unit, is a multiplicative cyclic group.

4. The method of claim 3, wherein, The step of the conference participant generating zero-knowledge proof verification information when attending the conference and the system verifying the identity of the conference participant according to the generated zero-knowledge proof verification information comprises: Constructing zero-knowledge proof circuits locally to a user Inputs, including public inputs and private inputs, are denoted as: ， wherein represents a hash value of Design constraint conditions, including: ， ， ， wherein, denotes a public key of a trusted unit, denotes a generator of denotes a bilinear map , is a multiplicative cyclic group, is a target multiplicative group; Finally, the zero-knowledge proof verification information is constructed, denoted as: ， In the formula, denotes a function that generates a cryptographic proof π; The system verifies the triple legality of the identity of the conference participant, including: Identity ownership: the user does indeed own the generation of the dynamic of the master private key ; Dynamic Legitimacy: Dynamic identity is based on , , , Correct generation; Identity authentication validity: the authentication signature held by the user is legal; the verification equation is wherein denotes the verification function.

5. The dynamic key threshold sharding privacy-tracing method based on blockchain collaborative verification according to any one of claims 1 to 4, characterized in that, The key information includes a conference ID, a screen watermark hash value, an encrypted salt value, an intermediate binding parameter, a hash value of a verifiable credential, a zero-knowledge proof, participant temporary virtual identity information and a constructed Merkle tree root value, an off-chain information storage address and timestamp information.

Images