Dynamic adaptive transaction method for interference data

Abstract

The application belongs to the field of data transaction, and provides a dynamic adaptive transaction method for interference data, which is applied to a corresponding system and is configured to: acquire a data block for transaction and negotiation constraint content in the data block transaction, so as to store transaction-related transaction funds into a preset smart contract and lock under the constraint of the negotiation constraint content; extract a sensitive feature summary in the data block based on a preset learning model, and generate a feature vector containing a sensitive coefficient according to the sensitive feature summary, so as to dynamically generate a noise value according to the feature vector; perform noise addition processing on the data block according to the noise value, and perform encryption on the noise-added data block according to a preset encryption algorithm; and verify the transaction information of the buyer and the owner and the encrypted data block respectively, so as to perform allocation of the transaction funds and decryption of the data block according to the verification result. The direct exposure of the original data is avoided, and the fairness of the data transaction is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data transaction, and particularly relates to a dynamic adaptive transaction method for interference data. BACKGROUND

[0002] At present, with the rapid development of digital economy, the demand for circulation and transaction of data as a key production factor continues to rise, but the traditional data transaction mode has contradictions that cannot be reconciled, which seriously restricts the efficient use of data elements.

[0003] In the traditional scheme, data transaction is mostly centered on the direct exchange of raw data. Although this mode allows buyers to obtain complete information without processing, it has significant drawbacks. On the one hand, raw data contains a large amount of sensitive information, and direct transaction can easily lead to privacy leakage risks. On the other hand, data owners are cautious about transactions due to concerns about raw data leakage or misuse, which makes it difficult for valuable data to enter the market.

[0004] At the same time, the traditional mode lacks effective data quality verification and transaction guarantee mechanisms. In existing data transaction schemes, the verification of data effectiveness mostly relies on smart contracts or zero-knowledge proofs, but this cannot be applied to all data. Some data effectiveness cannot be encoded as oracle functions in smart contracts, and it is also quite challenging to encode the correctness of data in zero-knowledge proof systems, and sometimes it is even impossible. After the transaction, if the data quality is found to be substandard, there is also a lack of fair and efficient dispute resolution approach, and data owners also face the risk of being refused payment after data delivery.

[0005] Therefore, there is an urgent need for a data transaction scheme that can avoid the direct exposure of raw data and ensure fair transaction. SUMMARY

[0006] The present application aims to at least solve the technical problems existing in the prior art, and provides a dynamic adaptive transaction method for interference data.

[0007] The application provides a dynamic adaptive transaction method for interference data, which is applied to a dynamic adaptive transaction system for interference data, and is characterized in that the dynamic adaptive transaction system for interference data is configured to perform the following steps: acquiring a data block for transaction and negotiation constraint content in the data block transaction, so that a purchaser and an owner of the data block store transaction funds related to the transaction into a preset smart contract and lock under the constraint of the negotiation constraint content; extracting a sensitive feature summary in the data block based on a preset learning model, and generating a feature vector containing a sensitive coefficient according to the sensitive feature summary, so as to dynamically generate a noise value according to the feature vector; performing noise processing on the data block according to the noise value, and encrypting the data block after noise processing according to a preset encryption algorithm; verifying the transaction information of the purchaser and the owner and the encrypted data block, so as to perform allocation of the transaction funds and decryption of the data block according to the verification result.

[0008] In an embodiment of the application, based on the above technical solution, the noise processing on the data block according to the noise value and the encryption of the data block after noise processing according to the preset encryption algorithm include: determining a data approximation degree of the data block after noise processing and the original data block according to the distribution of the data block, and calculating a privacy strength of the noise value according to a noise type of the noise value; when the data approximation degree and the privacy strength both satisfy a preset evaluation standard, the corresponding noise value is retained, and the data block after noise processing is generated.

[0009] In an embodiment of the application, based on the above technical solution, when the data approximation degree and the privacy strength both satisfy the preset evaluation standard, the corresponding noise value is retained, and the data block after noise processing is generated, including: acquiring the retained noise value, and generating a first noise vector according to a distribution type and an associated parameter corresponding to the noise value; calculating a second noise vector according to the first noise vector and a preset entanglement probability formula; expanding a noise pair composed of the first noise vector and the second noise vector based on a tensor product to obtain a noise matrix corresponding to the dimension of the data block; encrypting the noise matrix based on a preset elliptic curve encryption algorithm to generate the data block after noise processing.

[0010] In an embodiment of the present application, based on the above technical solution, the verification of the transaction information of the buyer and the owner respectively and the encrypted data block, and the execution of the allocation of the transaction fund and the decryption of the data block according to the verification result, when the verification result is failed and / or the buyer has objection, entering and executing a preset arbitration process to allocate the transaction fund according to the arbitration result output by the arbitration process; when the verification result is passed and the buyer has no objection, transferring the transaction fund to the allocation unit to allocate the transaction fund according to the negotiation constraint content.

[0011] In an embodiment of the present application, based on the above technical solution, when the verification result is failed and / or the buyer has objection, entering and executing a preset arbitration process to allocate the transaction fund according to the arbitration result output by the arbitration process, including: in the arbitration process, verifying the availability of the data block by a preset expert pool and making a voting decision; determining the fault party between the owner and the buyer according to the result of the voting decision, and executing a reward and punishment process according to a preset reward and punishment rule.

[0012] In an embodiment of the present application, based on the above technical solution, in the arbitration process, the availability of the data block is verified by a preset expert pool and a voting decision is made, including: the transaction parties respectively select a corresponding number of self-selected experts in a preset arbitration group according to a preset first proportion value; based on the smart contract, a corresponding number of random experts are randomly selected in a preset arbitration group according to a preset second proportion value; the self-selected experts and the random experts are combined to obtain a preset expert pool.

[0013] In an embodiment of the present application, based on the above technical solution, the transaction fund includes the purchase money of the buyer and the first deposit, and the second deposit of the owner; according to the result of the voting decision, the fault party between the owner and the buyer is determined, and a reward and punishment process is executed according to a preset reward and punishment rule, including: when the fault party is the buyer, the first deposit is allocated to the expert pool, and the second deposit and the purchase money are allocated to the owner; when the fault party is the owner, the second deposit is allocated to the expert pool, and the purchase money and the first deposit are allocated to the buyer.

[0014] In the technical solution of the embodiments of the present application, the above invention content can at least bring the following effects:

[0015] The application trains a feature vector through sensitive feature abstraction in a data block, and dynamically generates a noise value according to the feature vector, so as to encrypt the data block based on the noise value and a preset encryption algorithm, which can avoid direct exposure of the original data. Meanwhile, the transaction-related transaction funds are stored in a preset smart contract and locked under the constraint of negotiating the constraint content, and the encrypted data block and the obtained transaction information are verified, and the allocation of the transaction funds and the decryption of the data block are performed according to the verification result, which can improve the transaction fairness of the data transaction. It can be seen that the technical scheme of the application can avoid direct exposure of the original data while ensuring the transaction fairness of the data transaction.

[0016] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, and cannot limit the application. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 is a flow block diagram of a dynamic adaptive transaction method for interference data shown by an exemplary embodiment of the application;

[0018] Figure 2 is a schematic diagram of a Merkle hash tree shown by an exemplary embodiment of the application;

[0019] Figure 3 is an information flow direction schematic diagram of a data transaction system shown by an exemplary embodiment of the application;

[0020] Figure 4 is a flow chart of a data transaction shown by an exemplary embodiment of the application;

[0021] Figure 5 is a flow block diagram of an arbitration mechanism shown by an exemplary embodiment of the application. DETAILED DESCRIPTION

[0022] The embodiments of the application are described in detail below, and examples of the embodiments are shown in the drawings, wherein the same or similar notations represent the same or similar elements or elements with the same or similar functions throughout. The embodiments described below are exemplary and are only used to explain the application, and cannot be understood as a limitation of the application.

[0023] In addition, the described features, structures or characteristics can be combined in any suitable manner in one or more embodiments. In the following description, many specific details are provided to give a sufficient understanding of the embodiments of the application. However, one skilled in the art will realize that the technical solution of the application can be practiced without one or more of the specific details, or other methods, components, devices, steps, etc. can be used. In other cases, well-known methods, devices, implementations or operations are not shown or described in detail to avoid obscuring the aspects of the application.

[0024] The block diagrams shown in the drawings are merely functional entities and do not necessarily have to correspond to physically independent entities. That is, the functional entities can be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0025] The execution subject of the dynamic adaptive transaction method for interference-oriented data in the present application includes but is not limited to at least one of electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiments of the present application. In other words, the dynamic adaptive transaction method for interference-oriented data can be executed by software or hardware installed in a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to a single server, a server cluster, a cloud server, or a cloud server cluster, etc. The server can be a stand-alone server, or a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content distribution networks (CDN), and basic cloud computing services such as big data and artificial intelligence platforms, etc.

[0026] In the existing technical solution, data transactions are mostly targeted at raw data. Even if encrypted transmission is used, the raw data is still directly exposed to the data buyer after decryption, and there is a risk of misuse or secondary dissemination of sensitive information.

[0027] In addition, in the noise adding process, in order to ensure that the raw data of the data owner is not leaked, the noise adding process can only be operated by the data owner alone, and the data buyer cannot verify whether the noise strength meets the agreed privacy budget, which may result in over-noise and data invalidation. At the same time, there is also a lack of reliable verification means for whether the data block is tampered with by the data buyer in the noise verification process.

[0028] At the same time, the existing technical solution relies on automatic arbitration by encoding or centralized third-party arbitration, which has the application limitation that some domain data cannot be encoded, the risk of arbitration bias and untrustworthy third parties, and the dispute processing process is not transparent, which can easily lead to secondary disputes. In addition, the payment of funds relies on the credit of both parties, which may result in the situation that the data buyer refuses to pay or the data owner does not deliver the data after receiving the payment, and the trust cost is extremely high.

[0029] Figure 1 is a flowchart of the dynamic adaptive transaction method for interference-oriented data shown in the exemplary embodiments of the present application, as Figure 1As shown, in some embodiments of the present application, in order to reduce the data exposure probability in the data transaction process and improve the transaction fairness, the present application provides a dynamic adaptive transaction method for interference data, which is applied to a dynamic adaptive transaction system for interference data. The dynamic adaptive transaction system for interference data is configured to at least perform the following steps:

[0030] S100, acquiring a data block for transaction and negotiation constraint content in the data block transaction, so that the purchaser and the owner of the data block will deposit the transaction fund related to the transaction into the preset smart contract and lock under the constraint of the negotiation constraint content;

[0031] S110, extracting sensitive feature summary in the data block based on the preset learning model, and generating feature vector containing sensitive coefficient according to the sensitive feature summary, to dynamically generate noise value according to the feature vector;

[0032] S120, performing noise processing on the data block according to the noise value, and performing encryption on the data block after noise processing according to the preset encryption algorithm;

[0033] S130, respectively verifying the transaction information of the purchaser and the owner and the encrypted data block, to perform allocation of transaction fund and decryption of data block according to the verification result.

[0034] It should be noted that for the English description appearing in the full text, reference can be made to the explanation of the first occurrence.

[0035] Specifically, based on the above-mentioned dynamic adaptive transaction method for interference data, in the initial stage of data transaction, the data owner (DO) and the data purchaser (DP) need to negotiate the specific price (purchase money, the price required to purchase the data block), the guarantee deposit amount (first guarantee deposit, the guarantee deposit required to be paid by the purchaser) and (second guarantee deposit, the guarantee deposit required to be paid by the owner) of the transaction data, which usually satisfies , wherein The sensitive field classification system and the basic privacy constraint index are interactively negotiated. For example, in addition to the guarantee deposit, specific constraint conditions such as the specific range of use permission, the use scenario limitation under different permissions, etc.

[0036] For example, the sensitive field classification system includes a set of strong sensitive fields such as ID number, bank card number, etc., a set of medium sensitive fields such as mobile phone number, address, etc., a set of weak sensitive fields such as age, gender, etc., and a set of non-sensitive fields, and the basic privacy constraint indicators include a lower limit of privacy protection strength and a data utility threshold as boundary conditions for noise generation.

[0037] After the above negotiation of constraint content is agreed, the DP needs to transfer the transaction amount and the margin to the smart contract according to the negotiation constraint content of the smart contract (Smart Contract, SC), and the DO synchronously stores in the same SC. At this time, the SC enters a fund locking state, for example, its internal state variable status is set to pending, and only when the subsequent verification process passes and the DP has no objection to the transaction data, does not propose arbitration or meets the time constraint, the state transition to completed is triggered and the fund distribution is executed, or in the case of verification failure, it jumps to the refunded state. This mechanism ensures the traceability of the fund flow through the non-tamperable nature of the blockchain.

[0038] Of course, for the above example, the specific setting method of various state variables in the original control program for implementing the dynamic adaptive transaction method for interference data of the present application is only an exemplary reference to explain the explanatory description of part of the implementation mode of the present application, and is not specifically limited.

[0039] Further, it is illustrated that after the smart contract completes the locking of the transaction fund, the DO first divides the original data set Data into blocks according to a fixed block size, and obtains .

[0040] Then, for each data block , the DO needs to extract the corresponding sensitive feature summary , including dimension feature , sensitive distribution feature and attribute feature , that is .

[0041] Among them, the dimension feature is used to distinguish whether the data is high-dimensional or low-dimensional, the sensitive distribution feature represents the sensitive proportion vector, and the attribute feature indicates whether the data type is discrete or continuous.

[0042] The noise value generation process in this application's technical solution is illustrated by example. A reinforcement learning (RL) agent model, Model, is initialized through a pre-defined Trusted Collaborative Center (TCC). The agent's state space S is defined as a sensitive feature summary, i.e. Action space Adjusting noise distribution parameters, such as standard deviation. or scale parameter Increase or decrease, initial strategy The agent is then sent to the DO after random initialization or pre-trained model settings.

[0043] DO summarizes the previously extracted sensitive features ,use The agent model is locally trained to compute gradients, and these gradients are sent to the Trusted Collaboration Center (TCC). The TCC receives and aggregates these gradients to update the RL agent. Value function, ,in It's a reward.

[0044] Furthermore, TCC builds a simulation environment in which the RL agent undergoes multiple rounds of training, each round being called an episode.

[0045] TCC obtains the current state from the environment. , It includes dimensional features Sensitive distribution characteristics and attribute features The feature vector for the current state RL agent selects an action Specific adjustments are made to the noise distribution parameters, such as adjusting... The choice of actions follows strategy. Then, a noise vector is generated based on the adjusted parameters. For the original data block Simulated noise addition is performed to obtain the noise-added data. .

[0046] In some embodiments of the application, when adding noise to a data block based on a noise value, the following steps may be included at least:

[0047] Based on the distribution of data blocks, determine the data similarity between the noise-added data blocks and the original data blocks, and calculate the privacy strength of the noise values ​​based on the noise type of the noise values;

[0048] When both data approximation and privacy strength meet the preset evaluation criteria, the corresponding noise value is retained, and a noisy data block is generated.

[0049] Specifically, based on the noisy data blocks, TCC evaluates the noise-adding effect using two core metrics. The first metric is utility loss. The Kullback-Leibler divergence is used to measure the difference in distribution between the original data and the noisy data. The smaller the value, the closer the distribution of the data after noise addition is to the original data, and the better the utility is preserved.

[0050] The second metric is actual privacy budget. The actual differential privacy strength satisfied is calculated based on the noise type, such as for a Laplace distribution. The smaller the value, the stronger the privacy protection. Rewards are calculated based on the evaluation results. As a response to the current action Feedback ,in and For hyperparameters, This represents the minimum level of privacy.

[0051] Finally, based on the reward renew Value function, ,in It's the learning rate. It is to perform an action The new state that is entered later It is a discount factor.

[0052] The above steps will be repeated multiple times. After each round, the agent's... Value functions more accurately reflect the true value of a strategy. By gradually optimizing and iterating until the RL agent converges, the agent has learned a stable optimal strategy and can automatically select the optimal noise parameter adjustment scheme based on any data characteristics, thus achieving a balance between privacy and utility.

[0053] Through the above implementation methods, the technical solution of this application can be based on reinforcement learning-driven adaptive noise optimization. By extracting sensitive feature summaries of data blocks to construct the RL agent state space, and adjusting the noise distribution parameters as the action space, the trusted collaboration center conducts multiple rounds of reinforcement learning training to dynamically optimize the noise generation strategy, and finally achieves a precise balance between privacy protection strength and data utility.

[0054] It should be noted that in the traditional differential privacy data transaction, the noise generation is dependent on fixed parameters, which is easy to fall into the dilemma of insufficient privacy protection or data utility loss, and the noise generation is in an independent noise adding mode, without considering the correlation structure between the original data fields, which is easy to destroy the inherent correlation logic between the fields. To solve such technical problems, further, in some embodiments of the present application, when the data approximation degree and the privacy strength both satisfy the preset evaluation standard, the corresponding noise value is retained, and a noise-added data block is generated, which at least includes the following execution steps:

[0055] The retained noise value is obtained, and a first noise vector is generated according to the distribution type and the correlation parameter corresponding to the noise value;

[0056] The second noise vector is calculated according to the first noise vector and the preset entanglement probability formula;

[0057] The noise pair composed of the first noise vector and the second noise vector is expanded based on the tensor product to obtain a noise matrix corresponding to the dimension of the data block;

[0058] The noise matrix is encrypted based on the preset elliptic curve encryption algorithm to generate a noise-added data block.

[0059] It should be noted that the first noise vector in the above steps is represented by the following example , and the second noise vector is represented by .

[0060] Specifically, the following example is used for illustration, based on the learned optimization strategy , the TCC selects the final distribution type and parameters to generate the corresponding noise vector .

[0061] In order to retain the correlation structure information between the fields in the original data and reduce the utility loss caused by independent noise adding, especially for high-dimensional correlated data, the noise pair is used for noise adding, and the noise vector generated at this time is the principal component of the noise, which determines the overall noise strength.

[0062] After generating the principal component, the TCC calculates the Pearson correlation coefficient between the fields according to the field correlation degree of the sensitive distribution characteristics in the sensitive feature summary , and calculates the entanglement probability through a mapping function, wherein is the amplification coefficient.

[0063] Then the TCC generates a random number , if , then , otherwise .

[0064] Then generate the second random number , if , let , otherwise .

[0065] Finally, calculate the associated noise component , whose value depends on the principal component and the previously generated entangled bit pair , the calculation formula is , where is the correlation coefficient, , is the degree of violation of Bell's inequality, , is the rotated bit, is the random noise independent of , which ensures that is not completely determined by , and retains a certain randomness.

[0066] Based on the above steps, a pair of noise and can be obtained.

[0067] According to the different dimensions of the data, the two-dimensional noise pair and can be expanded to the corresponding dimension noise matrix by tensor product.

[0068] Based on the above embodiment, after generating the noise matrix , the TCC uses the elliptic curve encryption algorithm (ECC) to encrypt using the public key of DP, where is the elliptic curve base point, = , is the private key of DP, and is encrypted to obtain the ciphertext , where , is the order of , and finally the encrypted noise matrix is sent to the DO for data noise interference.

[0069] Through the above embodiments, the technical scheme of the present application can generate principal component noise and associated noise with correlation dependence based on the quantum entanglement noise generation mechanism, construct quantum entanglement constraint relationships by calculating field correlation and entanglement probability by mining the correlation structure characteristics between data fields, form a multi-dimensional noise matrix, effectively retain the correlation information between the original data fields while meeting the privacy budget, and avoid the loss of high-dimensional data utility caused by independent noise addition.

[0070] For the above explanation of the trusted collaborative center, the following explanation is made. The trusted collaborative center can be an exemplary virtual transaction center in the technical scheme of the present application to receive and process various data in the transaction process and output corresponding output results according to the specific processing process. In addition, the feature analysis model can be established based on one or more existing algorithm models. For example, in the embodiments of the present application, the feature analysis model can be established based on federated learning.

[0071] Among them, federated learning is a distributed machine learning framework that realizes joint modeling by allowing each participant to train the model locally and only exchange encrypted parameters. Without sharing the original data, privacy is effectively protected while model performance is improved.

[0072] Through the above manner, the present application extracts sensitive feature summaries in the data block and trains the extraction process based on a federated learning model to generate a feature vector containing a sensitive coefficient, selects different distribution types according to the vector, and dynamically generates different intensity noise to ensure that data usability is maximized while meeting the privacy budget.

[0073] Since the extraction of sensitive features is constantly changing with the results of model training, the present application can dynamically generate noise values based on the feature vectors of model training. And based on the noise value and the preset encryption algorithm, the data block is encrypted, so that the encrypted content is not easy to be disturbed, and the transaction security of the data block is improved.

[0074] Among them, the specific content of encrypting the data block according to the noise value and the preset encryption algorithm can be explained and described with reference to the following examples:

[0075] Specifically, the generation process of the noise value is explained as follows. The TCC of the present application selects different distribution types based on the feature vector Generates a noise vector Specifically, the TCC first selects different distribution types according to the dimension characteristics If it is high-dimensional data, Gaussian distribution is selected. The isotropic property of Gaussian distribution can ensure that the noise is uniformly diffused in the multi-dimensional space, and the standard deviation , The adjustment factor is determined by the basic privacy constraint index; for low-dimensional data, the Laplace distribution is chosen because low-dimensional data has simple field relationships, and the Laplace distribution is more computationally efficient. .

[0076] It should be noted that in traditional data transaction encryption schemes, key generation often adopts a fixed parameter mode, without considering the structural differences and sensitive complexity of different data blocks. To solve these technical problems, further, in some embodiments of this application, after generating noise values, the owner performs lattice-based encryption on the data block, which may include at least the following execution steps:

[0077] The sensitive feature vectors of the data block are transformed into high-dimensional point clouds, and the complexity features of the data block structure are extracted based on the high-dimensional point clouds according to the preset topology analysis.

[0078] The public key parameters of the encryption algorithm are adjusted according to the complexity characteristics to generate a key pair that matches the structure of the data block, and the data block is encrypted based on the key pair.

[0079] Specifically, this means that when DO receives the message sent by TCC... Afterwards, at least the following implementation steps need to be performed:

[0080] First, each data block Sensitive feature vectors It is considered as a high-dimensional point cloud, that is, each feature dimension corresponds to a coordinate axis in space, and the feature value corresponds to the coordinate value.

[0081] Then, the number of connected components of the point cloud is calculated using the Kruskal algorithm. The steps are as follows: calculate the pairwise distances of all points in the point cloud, connect the points in ascending order of distance, skipping any points where a connection would form a loop. The number of points in the final unconnected independent subset is the number of connected components. As a topological feature, it is used for dynamic adjustment of subsequent encryption parameters. The larger the value, the more complex the data block structure, and the stronger the encryption needs to be.

[0082] Next, a standard lattice basis matrix M is generated based on NTRU, which varies with topological features. Dynamically adjust the public key of lattice-based encryption , ,in This is a scaling factor to prevent the public key from becoming too large.

[0083] Private key Generated using the standard lattice basis reduction algorithm, ensuring compatibility with dynamic public keys. The private key must be able to efficiently decrypt content encrypted with the public key.

[0084] After generating the public-private key pair for lattice-based encryption, the data blocks... Mapped to binary polynomial Through public key Calculate lattice ciphertext This process leverages the computational difficulty of the lattice basis problem to ensure that a quantum computer cannot compute the solution in polynomial time. China Resumption This achieves quantum resistance.

[0085] Next, use DP's public key. ciphertext Homomorphic encryption is performed to obtain the encrypted double-layer ciphertext. Based on the additive homomorphism of the Paillier encryption algorithm, the encrypted double-layer ciphertext is processed within the ciphertext domain. Perform a noise-adding operation to obtain double-encrypted interference data. .

[0086] As can be seen, the technical solution disclosed in the above embodiments can generate dynamic keys based on topological data analysis. By converting data-sensitive feature vectors into high-dimensional point clouds and extracting data block structural complexity features using topological analysis, the public key parameters of lattice-based encryption are dynamically adjusted to generate lattice-based encryption key pairs that are compatible with the data structure. This achieves quantum attack protection while ensuring that the encryption strength matches the data-sensitive structural complexity.

[0087] Through the above implementation method, a two-layer encryption mechanism based on inner lattice-based encryption and outer homomorphic encryption is employed. First, the data owner uses lattice-based encryption to perform inner-layer encryption on the highly sensitive original data, and then homomorphic encryption is used for outer-layer processing to support noise addition to the ciphertext domain. This ensures the long-term privacy and security of highly sensitive data in the quantum era while preserving the efficiency and verifiability of data transactions, achieving a dual guarantee of quantum-resistant security and transaction practicality.

[0088] Meanwhile, through double-layer encryption and homomorphic characteristics, accurate verification of noise addition is achieved. The Trusted Collaboration Center verifies whether the encrypted original data block and the encrypted noise block are consistent with the encrypted interference data block in the ciphertext domain by using the encrypted original data block sent by the data owner and the encrypted interference data block sent by the data buyer, and by using the locally stored original interference noise. This mechanism directly verifies the compliance of the noise addition operation through cryptographic operations without leaking the original data.

[0089] in, Let be the modulus of the Paillier encryption algorithm, satisfying , and All are large prime numbers, and the final result is generated An encrypted jamming data block, i.e. , the DO will encrypt the interference data block with the DP public key . and send it to the DP.

[0090] Based on this, the application can block the exposure of original data from the source through the noise adding mechanism. The data owner only transmits the encrypted interference data block to the data buyer, and the data buyer can only obtain the noise-added data after decryption, and cannot restore the original information, which solves the problem of privacy leakage in direct transaction of original data to a certain extent.

[0091] Meanwhile, in the technical scheme of the application, the inner layer of the high-sensitive data block is encrypted by lattice to realize quantum resistance, and the outer layer uses homomorphic encryption algorithm to ensure operation in ciphertext state, and the cooperation of the double-layer encryption algorithm forms the advantages of long-term security and privacy.

[0092] Further exemplary description, in the technical scheme of the application, after the DP receives the encrypted interference data block sent by the DO, the private key of the DP is used to perform decryption operation, and the lattice ciphertext can be obtained. At this time, the smart contract triggers the temporary authorization mechanism, and the DP can temporarily obtain the use permission of to calculate , that is, the noise-added available data required by the DP.

[0093] In some embodiments of the application, based on the above embodiments, Figure 2 is a schematic diagram of the Merkle hash tree shown in the exemplary embodiments of the application, as shown in Figure 2 , the DO constructs a Merkle hash tree (MHT) with each as a leaf node while sending to the DP.

[0094] Specifically, the DO first calculates the hash value of each leaf node , and then recursively calculates the parent node hash in layers, that is, , where represents the index of the node in the layer, starting from 0, represents the layer level of the node, is the leaf layer, is the layer above the leaf layer, and so on, such as the hash value of the first node in the leaf layer , and the Merkle hash tree is constructed through the above calculation until the root hash is generated, where represents the number of nodes.

[0095] After the construction is completed, the DO will send the transaction information to the DP. The hash path corresponding to each node contains all the intermediate node hashes from to and is sent to the TCC for subsequent integrity verification.

[0096] Based on this, the TCC generates random indexes using a cryptographically secure pseudo-random number generator after receiving the MHT information sent by the DO, and sends to the DP and the DO, respectively.

[0097] Illustratively, in the specific technical solution of the present application, the DP needs to extract the corresponding encrypted block from according to the index S. It should be noted that here is the transaction initial DO sent to the DP, so the DP does not need to perform any encryption or other operations here, but only needs to select the corresponding encrypted interference data block according to the index, which also improves the efficiency of the entire transaction. Finally, the DP forwards the selected to the TCC for verification.

[0098] Based on the above embodiment, for the DO, the corresponding original data block needs to be calculated according to the received index using the lattice public key to obtain the lattice ciphertext , then encrypted using the public key of the DP to obtain , , and then encrypted using its own public key to obtain , , and finally send the encrypted original data block to the TCC.

[0099] After encrypting the data block, the technical solution of the present application verifies the encrypted data block and the obtained transaction information, and transfers the transaction funds to the distribution unit to perform the distribution of the transaction funds and the decryption of the data block according to the verification result.

[0100] Specifically, after receiving the encrypted interference data block sent by the DP, the TCC first performs MHT verification on sent by the DP.

[0101] The TCC first calculates the hash value of the encrypted interference data block to be verified using the same hash algorithm as the DO, i.e. , denoted as , then extract the path information of the corresponding encrypted interference data block from the hash path provided by the DO, and then calculate the parent node hash layer by layer upwards according to the hash path to finally obtain a derived root hash , starting from the hash path, calculate the parent node hash layer by layer upwards, and finally obtain a derived root hash , if is equal to the original Merkle root sent by the DO , it means that the encrypted interference data block sent by the DP is not tampered with, which is the original data block, otherwise, the data block is tampered with, triggering the transaction termination process.

[0102] If the verification is passed, that is, the set by the program is executed

[0103] First, the TCC encrypts the encrypted interference data block sent by the DP after verification using the public key of the DO to obtain , .

[0104] Then the TCC uses the noise corresponding to the index generated by itself at the beginning to perform the same encryption, it should be noted that the noise used here is saved in the TCC, and the DO cannot be forged to pass the verification.

[0105] It can be seen that the noise of the present application is dynamically generated by an independent third-party trusted collaboration center based on data characteristics, and the noise is encrypted by an encryption algorithm and sent to the data owner for noise addition. Not only ensures the compliance of noise generation, but also the data owner can only add noise to the data in the ciphertext domain, and cannot access the original noise value.

[0106] Then use the public key of the DP to perform the first encryption, obtain , , and then use the public key of the DO to perform the second encryption, obtain , .

[0107] Finally, use the encrypted original data block sent by the DO , to make the following calculation in the ciphertext domain:

[0108]

[0109] Based on the above calculation formula, after the calculation is completed, the TCC generates a set of random numbers , Then use the multiplication homomorphism of the Paillier encryption algorithm to calculate the verification result , if the encrypted zero value, then the proof , i.e. the DO noise addition is correct, the transaction can continue, otherwise it means that the DO added more noise than the TCC generated to better protect its original data privacy, at this time the SC automatically performs a refund operation, returns the Price and to the DP, to the DO, and the transaction is terminated.

[0110] It can be seen that the trusted coordination center of the present application verifies part of the data blocks by randomly generating an index, calculates the hash value of the extracted data blocks using Merkle proof, and then derives the root hash layer by layer according to the hash path. If it is consistent with the root hash value provided by the data owner, it means that the sampling verification data block sent by the data buyer is correct and has not been tampered with, which greatly reduces the calculation cost and time cost of verification.

[0111] Through the above implementation, the dynamic noise generation mechanism based on data feature perception analyzes the distribution characteristics of the original data through federated learning, and then generates noise that is adapted to the data characteristics. This can not only ensure that high-sensitive data obtains stronger privacy protection, but also avoid low-sensitive data losing statistical utility due to excessive noise, significantly improving the practicality of differential privacy technology in data transaction scenarios.

[0112] In some embodiments of the present application, after verifying the transaction information of the buyer and the owner respectively, and the encrypted data blocks, the allocation of transaction funds and the decryption of data blocks are executed according to the verification results. At least the following execution steps can be included:

[0113] When the verification result is not passed and / or the buyer has objections, enter and execute the preset arbitration process to allocate the transaction funds according to the arbitration result output by the arbitration process;

[0114] When the verification result is passed and the buyer has no objections, transfer the transaction funds to the allocation unit to allocate the transaction funds according to the negotiation constraint content.

[0115] Specifically, based on the content of the above embodiments, the following further explanations are made. In the above verification process, the data blocks in the technical solution of the present application need to judge whether the noise added to the encrypted data blocks is correct after adding the noise value;

[0116] When the judgment result is correct, the transaction process is continued, otherwise the transaction process is terminated.

[0117] Specifically, through the sampling inspection of noise value addition by the trusted coordination center, the trusted coordination center performs interference calculation on the sampled original data block encrypted by the public key of both parties and the encrypted noise, verifies whether the noise addition result meets the expectation, enforces noise addition compliance, and avoids malicious manipulation of data quality by the data owner.

[0118] It can be seen that the technical solution of the present application is based on transaction interference data, not original data, which reduces the risk of privacy leakage from the source, and at the same time realizes the transaction process verifiable and the dispute fair and just by means of technology, providing an effective and innovative solution for the safe circulation of data.

[0119] For example, if the interference noise added by the DO is correct and the DP has no objection to the availability of the data, the DP needs to send a confirmation signal confirm to the SC within the time-lock time, such as 24 hours, which contains a digital signature Signed with a private key to prevent malicious forgery of the confirmation signal.

[0120] After receiving and verifying, the SC will automatically send the pre-stored transaction amount to the account address of the DO, and return the deposit and of the DO and the DP to the account addresses of both parties.

[0121] If the SC does not receive from the DP within the time-lock time, it is assumed that the DP has no objection to the availability of the data, otherwise the DP can propose arbitration, so that the SC will automatically perform the same operation as above when is not received within the time-lock time.

[0122] In some embodiments of the present application, based on the above embodiments, when the verification result is not passed and / or the buyer has an objection, a preset arbitration process is entered and executed to allocate the transaction funds according to the arbitration result output by the arbitration process, which can at least include the following execution steps:

[0123] In the arbitration process, the availability of the data block is verified by a preset expert pool and a voting decision is made.

[0124] According to the result of the voting decision, the wrong party among the owner and the buyer is determined, and a reward and punishment process is executed according to the preset reward and punishment rules.

[0125] Specifically, if the DP has objection to the data availability and proposes arbitration, the arbitration process is started. The application introduces a preset expert pool (EP) for different field data. The experts in the expert pool can judge the quality of the data. According to the size of the transaction data, the number K of experts of the arbitration panel (AP) is set. The larger the size of the transaction data, the larger the value of K. When determining the experts of the arbitration panel, the DP selects K experts from the EP , the DO selects K experts from the EP, and the remaining K experts are randomly selected by the SC through a random selection algorithm to finally form the arbitration panel.

[0126] At the same time, in order to ensure that different experts do not interfere with each other when voting, and to ensure the independence and fairness of the voting, the arbitration stage adopts ring signature (Ring Signature) technology to realize anonymous voting.

[0127] Each expert has a key pair , , and the ring member set is .

[0128] Each expert independently votes according to the inspection of the data quality, and the voting result is , where 0 represents that the data is unavailable, and 1 represents that the data is available. Finally, the voting result is signed to obtain . .

[0129] Exemplarily, based on the above setting of the expert pool, when a dispute occurs, each party selects 1 / 3 field experts, and the remaining 1 / 3 is randomly selected to form an arbitration panel. Through ring signature technology, anonymous voting is realized, and the result is stored on the chain for evidence.

[0130] Through the above implementation, through anonymous voting, interference during voting is avoided, random selection ensures neutrality, and the smart contract automatically executes the arbitration result, realizing the fairness and transparency of dispute resolution.

[0131] According to the voting result of the above expert pool, the fault party among the owner and the buyer is determined, and the reward and punishment process is executed according to the preset reward and punishment rules.

[0132] Specifically, the transaction fund includes the purchase money of the buyer and the first margin, and the second margin of the owner. Of course, the first margin and the second margin here are only for convenient description, and their specific contents can be understood as corresponding to the above-mentioned purchase money margin and the owner's margin.

[0133] Further, according to the result of the voting decision, the wrong party among the owner and the buyer is determined, and a reward and punishment process is performed according to a preset reward and punishment rule, which at least includes the following execution steps:

[0134] When the wrong party is the buyer, the first deposit is allocated to the expert pool, and the second deposit and the purchase money are allocated to the owner;

[0135] When the wrong party is the owner, the second deposit is allocated to the expert pool, and the purchase money and the first deposit are allocated to the buyer.

[0136] Based on the above embodiments, exemplary description is made, all signatures After being submitted to the SC, the SC will first verify each Whether the consistency condition of the ring signature is met, that is, the validity of the signature is confirmed, and the votes are counted.

[0137] If , it is proved that the data is unavailable, and it is determined that the DO is dishonest, and the SC will automatically deduct the deposit of the DO pre-stored, and divide it into the experts in the AP as an arbitration reward, and return the transaction amount and the deposit pre-stored by the DP to the account address of the DP.

[0138] On the contrary, if , it is proved that the data is available, which means that the DO wants to get the data without paying the amount, and there is dishonest behavior, so the SC will automatically deduct the deposit of the DP, and divide it into the experts in the AP as an arbitration reward, and send the transaction amount and the deposit of the DO to the account address of the DO.

[0139] It can be seen that the technical scheme of the present application introduces a deposit mechanism, which requires both parties of the transaction to pre-store a deposit according to a certain percentage of the transaction price, and locks the transaction amount and the deposits of both parties through a smart contract. With the automatic execution and non-tamperability of the blockchain smart contract, the safety and traceability of the fund flow are ensured, and the existence of the deposit improves the honest willingness of both parties of the transaction.

[0140] It should be noted that the arbitration stage of the present application is not necessary, but optional, and under the arbitration mechanism of the present application, any dishonest party will suffer a considerable loss, so the probability of entering the arbitration stage is small. The arbitration stage is only an optional scheme to improve the transaction efficiency.

[0141] Through the above embodiments, the two parties before the transaction will deposit the margin and the transaction amount into the contract, and only after the data verification is passed and there is no dispute or dispute ruling, the smart contract will automatically complete the transfer or return, replacing credit guarantee with technical means, reducing human intervention.

[0142] Figure 3 is a schematic diagram of the information flow of the data transaction system shown in the exemplary embodiments of the present application, as shown in Figure 3 To disclose the technical solutions of the present application more clearly, in combination with the specific content of the above embodiments, the following further explanations are made:

[0143] The technical solutions of the present application break through the limitations of traditional raw data transactions, and innovatively take interference data as the core of the transaction, combined with federated learning, smart contract, differential privacy, homomorphic encryption, lattice encryption, Merkle proof and ring signature, etc. to build a transaction mechanism that gives equal weight to privacy protection and fair protection in the whole process.

[0144] Specifically, Figure 4 is a flowchart of the data transaction shown in the exemplary embodiments of the present application, as shown in Figure 4 For the data transaction process in the technical solutions of the present application, at least the following execution steps can be included:

[0145] 400. At the initial stage of the transaction, the data owner and the data buyer negotiate to determine the data price, the margin, the sensitive field classification system and the basic privacy constraint index, and the two parties deposit the margin and the transaction amount into the smart contract to lock the funds to ensure the safety of the transaction.

[0146] 410. After the smart contract funds are locked, the data owner divides the raw data into blocks and extracts the sensitive feature abstract, and the trusted collaboration center initializes the reinforcement learning agent model, and optimizes the noise generation strategy after multiple rounds of training.

[0147] 420. Combined with the quantum entanglement characteristics, an associated noise matrix is generated and encrypted and sent to the data owner.

[0148] 430. The data owner dynamically adjusts the lattice encryption parameters through topological data analysis, uses lattice encryption to protect the data to achieve quantum resistance, then performs homomorphic encryption, adds noise in the ciphertext state after encryption, generates encrypted interference data blocks and sends them to the data buyer. Through the double means of noise addition and encryption, privacy leakage is avoided, and the core privacy pain point of traditional raw data transactions is solved.

[0149] 440、To ensure data integrity and correctness of noise addition, the application also introduces a Merkle hash tree and a sampling verification mechanism. The data owner generates a Merkle hash tree with encrypted interference data blocks as leaf nodes, submits the root value and hash path to the trusted coordination center, the trusted coordination center randomly selects the index value of the sampling inspection and sends it to both parties, the data buyer submits the encrypted interference data block corresponding to the index, the data owner submits the encrypted original data block corresponding to the index, and the trusted coordination center verifies that the data has not been tampered with through Merkle proof, and further verifies whether the data owner has honestly added noise.

[0150] 450、If the noise addition is correct and the data buyer has no objection to the data availability, a confirmation signal is sent to the smart contract, and the smart contract transfers the transaction amount to the owner within the agreed time and returns the deposit of both parties.

[0151] Figure 5 is a flow chart of the arbitration mechanism shown in the exemplary embodiments of the application, as shown in Figure 5 based on the above transaction process, the application also proposes an arbitration mechanism. If the data buyer has no objection to the data availability, the smart contract transfers the transaction amount to the owner within the agreed time and returns the deposit of both parties.

[0152] 500、If there is an objection, arbitration is started, and each party selects 1 / 3 of the arbitrators from the pre-set pool of domain experts, and the remaining 1 / 3 is randomly selected to form an arbitration group.

[0153] 510、The arbitration group verifies the data availability and votes, and in order to prevent mutual interference, anonymous voting is carried out using ring signature technology, and finally the voting results are submitted to the smart contract.

[0154] 520、According to the voting results, if the data quality of the data owner is low, the smart contract deducts the deposit of the data owner to the arbitration group as a reward, and returns the transaction amount and the deposit of the data buyer.

[0155] 530、If the data buyer is not honest and maliciously objects, the deposit of the data buyer is deducted to the arbitration group as a reward, the deposit of the data owner is returned, and the transaction amount is given to the data owner.

[0156] The arbitration group determines the responsibility through ring signature anonymous voting, and if the data quality is not up to standard, the deposit of the data owner is deducted to reward the arbitrators, and the funds of the data buyer are returned. If the data buyer maliciously objects, his deposit is deducted, and the payment to the owner is completed.

[0157] It can be seen that the technical scheme of the application introduces a dynamic arbitration and ring signature voting mechanism, the arbitration group is composed of one-third of experts selected by the data owner, one-third of experts selected by the data buyer and one-third of experts randomly selected by the smart contract, the experts vote anonymously through ring signature, the independence and fairness of the voting are ensured, the smart contract automatically executes rewards and punishments according to the voting results, and human intervention is reduced.

[0158] It should be noted that the dynamic adaptive transaction system for interference data provided in the embodiment belongs to the same concept as the dynamic adaptive transaction method for interference data provided in the above embodiment, and the specific manner in which each module and unit performs operations has been described in detail in the method embodiment, which will not be repeated here. The dynamic adaptive transaction system for interference data provided in the embodiment can complete the above-described functions by different functional modules according to the needs in actual application, that is, the internal structure of the device is divided into different functional modules to complete all or part of the above-described functions, and this is not limited herein.

[0159] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "example", "specific example", "one implementation", "one preferred implementation" or "some examples" and the like means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0160] Although the embodiments of the present application have been shown and described, those skilled in the art can understand that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and purposes of the present application, and the scope of the present application is defined by the claims and their equivalents.​​​

Claims

1. A method for dynamic adaptive transaction for interference data, applied to a dynamic adaptive transaction system for interference data, characterized in that, The dynamic adaptive transaction system facing interference data is configured to perform the following steps: According to the negotiation constraint content in the data transaction, the data buyer and the data owner deposit transaction funds related to the transaction into a preset smart contract and lock under the constraint of the negotiation constraint content; Acquiring data blocks for transaction , is an original data set After blocking, the first data block, based on the preset learning model, extracts the sensitive feature summary in the data block , and generates a feature vector containing a sensitive coefficient according to the sensitive feature summary, to dynamically generate a noise value according to the feature vector; According to the noise value, the data block is subjected to noise adding processing, and the data block after the noise adding processing is subjected to encryption according to a preset encryption algorithm. According to the distribution of the data block , determine the data approximation degree of the data block with the original data block , and calculate the privacy strength of the noise value according to the noise type of the noise value; When the data approximation degree and the privacy intensity both satisfy a preset evaluation standard, a corresponding noise value is reserved; A first noise vector is generated according to a distribution type and associated parameters of the reserved noise value; A second noise vector is calculated according to the first noise vector and a preset entanglement probability formula; expanding a noise pair composed of the first noise vector and the second noise vector based on a tensor product to obtain the data block a noise matrix of corresponding dimensions; The noise matrix is encrypted based on a preset elliptic curve encryption algorithm to generate an encrypted noise matrix; transforming the sensitive feature vector of the data block into a high-dimensional point cloud, to extract the complexity feature of the data block structure based on the high-dimensional point cloud according to a preset topology analysis. adjusting public key parameters of a lattice encryption algorithm according to the complexity feature to generate a lattice encryption algorithm with a public key key pairs corresponding to the structure adaptation; According to the data block and the public key of the key pair to compute a lattice ciphertext; An encrypted data block is obtained according to the lattice ciphertext, the Paillier encryption algorithm and the encrypted noise matrix; The transaction information of the buyer and the owner and the encrypted data block are verified respectively, and the allocation of the transaction funds and the decryption of the data block are performed according to the verification result.

2. The dynamic adaptive transaction method for interference data according to claim 1, wherein, The verification of the transaction information of the buyer and the owner and the encrypted data block respectively, and the allocation of the transaction funds and the decryption of the data block according to the verification result, comprises: When the verification result is not passed and / or the buyer has objections, a preset arbitration process is entered and executed, and the transaction funds are allocated according to the arbitration result output by the arbitration process; When the verification result is passed and the buyer has no objections, the transaction funds are transferred to an allocation unit, and the transaction funds are allocated according to the negotiation constraint content.

3. The dynamic adaptive transaction method against interference data according to claim 2, characterized in that, When the verification result is not passed and / or the buyer has objections, a preset arbitration process is entered and executed, and the transaction funds are allocated according to the arbitration result output by the arbitration process, comprising: In the arbitration process, the availability of the data block is verified by a preset expert pool and a voting decision is made; According to the result of the voting decision, the wrong party among the owner and the buyer is determined, and a reward and punishment process is performed according to a preset reward and punishment rule.

4. The dynamic adaptive transaction method against interference data according to claim 3, characterized in that, In the arbitration process, the availability of the data block is verified by a preset expert pool and a voting decision is made, comprising: The transaction parties respectively select a corresponding number of self-selected experts in a preset arbitration group according to a preset first proportion value; Based on the smart contract, a corresponding number of random experts are randomly selected in a preset arbitration group according to a preset second proportion value; The self-selected experts and the random experts are combined to obtain a preset expert pool.

5. The dynamic adaptive transaction method for interference data according to claim 3, wherein, The transaction funds include the purchase money and the first deposit of the buyer, and the second deposit of the owner; according to the result of the voting decision, the wrong party among the owner and the buyer is determined, and a reward and punishment process is performed according to a preset reward and punishment rule, comprising: when the at fault party is the purchaser, allocating the first deposit to the pool of experts and allocating the second deposit and the purchase money to the owner; when the at fault party is the owner, allocating the second deposit to the pool of experts and allocating the purchase money and the first deposit to the purchaser.

Images