Data acquisition method and system
By managing the credibility of sensing devices through blockchain and encryption technology, verifiable noisy data is generated, solving the problem of noise affecting data accuracy in existing technologies and ensuring privacy protection and data quality.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-14
- Publication Date
- 2026-03-27
AI Technical Summary
In existing technologies, adding noise to protect privacy using differential privacy techniques can affect the accuracy of data aggregation and lead to a decline in data quality.
By employing blockchain technology and encryption methods, verifiable noisy data is generated through the reputation management and encryption/decryption process of sensing devices. The data is then verified based on target weights and trusted intervals to ensure data quality and privacy protection.
End-to-end privacy protection is achieved to prevent the leakage of participant privacy, and the final target perception data is free of noise, ensuring data quality and security.
Smart Images
Figure CN121744374A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of data privacy, in particular to a data collection method and system. BACKGROUND
[0002] Crowd sensing is a sensing task performed by intelligent sensing devices to collect data. In the prior art, differential privacy technology is usually used to protect privacy by adding noise, but the added noise affects the accuracy of the final data aggregation, resulting in a decrease in data quality. SUMMARY
[0003] To solve the above technical problems, the present application provides a data collection method and system that can ensure that the added noise is verifiable and improve data quality.
[0004] Specifically, the present application provides a data collection method, comprising the following steps: obtaining encrypted first noise-added data synthesized on a blockchain according to initial sensing data, and decrypting the first noise-added data to obtain second noise-added data; calculating a target weight based on the second noise-added data, and calculating a target data true value based on the target weight and the initial sensing data; and constructing a trusted interval based on the target data true value, verifying the initial sensing data based on the trusted interval, and taking the initial sensing data that passes the verification as target sensing data.
[0005] In the above technical solution, the data requester and the aggregation server cannot obtain the original sensing data, which fundamentally prevents the leakage of participant privacy, realizes end-to-end privacy protection, and the final target sensing data is not noise-added, which guarantees the quality of the data.
[0006] Further, the initial sensing data is derived from a plurality of sensing devices; the obtaining of the encrypted first noise-added data synthesized on the blockchain according to the initial sensing data comprises: determining a plurality of sensing devices according to the credibility of the sensing devices, obtaining initial sensing data through the sensing devices, and calculating sensing data ciphertext based on a first preset random number, a first public key and the initial sensing data through the sensing devices; and obtaining noise ciphertext calculated and generated based on the first preset random number, the first public key and a preset noise value, to synthesize noisy data on the blockchain according to the sensing data ciphertext and the noise ciphertext.
[0007] In the above technical solution, only the sensing device can obtain the corresponding plaintext data from the sensing data ciphertext through its own private key, and other entities cannot infer the value of the private key, so they cannot steal the sensing data; the tamper-proof nature of the blockchain ensures the security and integrity of the noisy data during the synthesis process, preventing malicious tampering of the data.
[0008] Further, the encrypted first noisy data synthesized according to the initial sensing data on the blockchain further includes: re-encrypting the noisy data based on a re-encryption key as the encrypted first noisy data; wherein the re-encryption key is generated by the sensing device based on a second public key and a first private key.
[0009] In the above technical solution, the privacy protection of data is further enhanced, making data more secure during transmission and storage.
[0010] Further, the obtaining of the second noisy data includes: decrypting the first noisy data using a second private key to obtain the second noisy data.
[0011] In the above technical solution, the availability of data is ensured, and the decryption process still relies on the private key, maintaining the security of the data.
[0012] Further, the obtaining of the target weight according to the second noisy data includes: calculating an initial true value according to the second noisy data, obtaining an initial weight based on the initial true value and iteratively updating; wherein when the absolute value of the difference between the latest initial true value and the initial true value updated last time is less than a set threshold, the initial weight corresponding to the latest initial true value is taken as the target weight.
[0013] In the above technical solution, the iterative updating method can continuously optimize the weight, so that the target weight can more accurately reflect the true situation of the data.
[0014] Further, the obtaining of the target data true value according to the target weight and the initial sensing data includes: obtaining a first true value based on a first public key and a random interference noise vector, and obtaining an interference noise ratio according to the first true value and the sensing data ciphertext; re-encrypting the interference noise ratio based on a re-encryption key, and decrypting the re-encrypted interference noise ratio based on the re-encryption key to obtain a noisy plaintext vector; and obtaining a second true value based on a second public key and a random interference noise vector, and decrypting the second true value based on a second private key to obtain a target interference noise vector, while sampling a random mask scalar.
[0015] In the above technical solution, the noise in the data can be quantified and processed by calculating the interference noise ratio, providing a basis for subsequent noise removal; the re-encryption process ensures the security of the interference noise ratio during processing, and the noisy plaintext vector obtained after decryption can be used for subsequent calculation; the introduction of the random mask scalar can further protect the privacy of the data and prevent the data from being leaked during calculation.
[0016] Further, the calculating the target data true value according to the target weight and the initial perception data further comprises: obtaining a mask weighting value and an unmasked weight sum according to the target weight, respectively encrypting the mask weighting value and the unmasked weight sum based on a second public key; obtaining a mask aggregated true value encrypted based on the second public key on a blockchain; wherein the mask aggregated true value is obtained by respectively decrypting the mask weighting value and the unmasked weight sum based on a second private key on the blockchain and combining the random mask scalar; and decrypting the encrypted mask aggregated true value based on the second private key and removing the random mask scalar to obtain the target data true value.
[0017] In the above technical solution, through a series of encryption, decryption and calculation operations, the target data true value obtained finally can effectively remove the influence of noise and improve the quality of data.
[0018] Further, the constructing a trusted interval based on the target data true value to verify the initial perception data based on the trusted interval comprises: constructing a trusted interval based on the target data true value and a set error; encrypting the initial perception data based on a first public key by the perception device to obtain homomorphic ciphertext, and constructing an affine function according to a second preset random number and a preset offset value to obtain target ciphertext and a trusted end point based on the affine function, homomorphic ciphertext and the trusted interval; decrypting the target ciphertext based on a first private key by the perception device to compare the decrypted target ciphertext with the trusted end point; wherein it is judged whether the target ciphertext is within the interval corresponding to the trusted end point based on the second preset random number, if yes, it is determined that the verification is passed; otherwise, it is determined that the verification fails, and the corresponding initial perception data is rejected.
[0019] In the above technical solution, the encryption process ensures the privacy of data in the processing process, and the construction of the affine function can further transform and process the data; through data verification, data with higher quality is effectively screened out, and the reliability of the final target perception data is improved.
[0020] Further, it further comprises: updating the reputation of the perception device based on the verification result.
[0021] In the above technical solution, the perception device with high reputation may be used more in the subsequent data acquisition process, and the perception device with low reputation may be reduced in use or eliminated. This incentive mechanism can encourage the perception device to provide more accurate and reliable data and improve the performance of the entire data acquisition system.
[0022] Based on the same concept, the application also provides a data collection system, comprising: an acquisition module, configured to acquire encrypted first noisy data synthesized from initial sensing data on a blockchain, and decrypt the first noisy data to obtain second noisy data; a calculation module, configured to calculate a target weight based on the second noisy data, and calculate target data true value based on the target weight and the initial sensing data; and a verification module, configured to construct a trusted interval based on the target data true value, to verify the initial sensing data based on the trusted interval, and take the initial sensing data that passes the verification as target sensing data.
[0023] In the above technical solution, the acquisition module ensures the security and availability of data, providing a basis for subsequent calculation and verification; the calculation module improves the accuracy and quality of data, effectively removing the influence of noise; and the verification module selects data with high quality as target sensing data, ensuring the reliability of data.
[0024] Compared with the prior art, the application has the beneficial effects that: The application can realize that neither the data requester nor the aggregation server can obtain the original sensing data, fundamentally preventing the leakage of participant privacy, realizing end-to-end privacy protection, and finally obtaining target sensing data without noise, ensuring the quality of data. BRIEF DESCRIPTION OF DRAWINGS
[0025] Figure 1 The flowchart of the data collection method described in the application.
[0026] Figure 2 The framework diagram of the data collection system described in the application.
[0027] Figure 3 The average absolute error comparison diagram under different malicious worker proportions described in the application.
[0028] Figure 4 The root mean square error comparison diagram under different malicious worker proportions described in the application.
[0029] Figure 5 The worker reputation change diagram described in the application. DETAILED DESCRIPTION
[0030] The data collection method and system of the application will be further described in detail below in combination with specific embodiments and drawings.
[0031] Please refer to Figure 1 The application provides a data collection method, comprising the following steps S100-S300.
[0032] In one feasible implementation, a data requester issues a task, workers equipped with sensing devices select the task, and the data requester then selects workers based on their reputation and instructs them to perform the sensing task, thereby obtaining sensing data. Noise is then generated by perturbing the server, and encrypted noisy data is synthesized on the blockchain. Finally, the data requester decrypts the noisy data to obtain noisy data according to agreed-upon rules. Furthermore, after calculating weights using the noisy data, the data requester uses these weights to calculate the truth value using noise-free data on the blockchain while maintaining privacy. Finally, based on the calculated truth value, the data requester sets a reasonable range for data values, decreasing the reputation of workers whose reported data falls outside this range and increasing the reputation of those within the range. By dynamically evaluating worker reputation, trustworthy workers are selected, thus ensuring data quality at the source.
[0033] The specific implementation methods of steps S100-S300 will be described in detail below.
[0034] Step S100: Obtain the encrypted first noisy data synthesized on the blockchain based on the initial perception data, and decrypt the first noisy data to obtain the second noisy data.
[0035] The initial sensing data originates from multiple sensing devices. The process of acquiring the encrypted first noisy data synthesized on the blockchain based on the initial sensing data includes: determining multiple sensing devices based on their reputation, obtaining the initial sensing data through these devices, and calculating the sensing data ciphertext using the sensing devices based on a first preset random number, a first public key, and the initial sensing data; and acquiring the noise ciphertext generated based on the first preset random number, the first public key, and a preset noise value, to synthesize noisy data on the blockchain based on the sensing data ciphertext and the noise ciphertext.
[0036] In some embodiments, data requesters and workers Each user registers on the blockchain, and the trusted center assigns a public-private key pair to the data requester. At the same time, for workers Also allocate a public-private key pair ;in, , These represent the second public key and the first public key, respectively. , These are represented as the second private key and the first private key, respectively.
[0037] The key generation process is as follows: The selection order is a positive integer. Multiplication cyclic group and , making existence bilinear mapping ,in The generator is From the non-degeneracy of bilinear mappings, we can know that The generator is At this time, the workers Select a random positive integer As a private key, i.e. This allows workers to generate their public keys. Similarly, it can be deduced that the private key of the data requester is... Then its public key can be represented as .
[0038] It should be noted that a data requester (DR) is someone who needs data and publishes requirements such as the location and time of the data sensing task; a worker refers to a person carrying sensing devices who can perform the task of sensing data; and a blockchain is a decentralized distributed ledger that is block-based, immutable, secure, and reliable. It combines distributed storage, peer-to-peer transmission, consensus mechanisms, cryptography, and other technologies to record transactions and information through a continuously growing chain of data blocks, thereby ensuring data security and transparency.
[0039] Furthermore, data requesters publish tasks on the blockchain, workers search for tasks on the blockchain, and if a worker is willing to do a task, they apply for the task from the data requester; then the data requester selects workers based on their creditworthiness. Workers are selected in the following way: If it falls The workers in the middle will choose one to land on. The workers in the middle, and then from Select the workers with the highest level of trust from the set. One worker; if not landed The workers in the middle, from Select the set with the highest trust level One worker.
[0040] It should be noted that the data requester pre-divided the workers into three groups. This is a set of trustworthy workers, initially empty; when a worker's trustworthiness exceeds a predetermined threshold... Time-return set ; Let the set of workers with undetermined trust levels initially consist of all workers, and the trust levels of the workers in this set be... ; This is a set of malicious workers, initially empty. If a worker's reputation is lower than... When workers are included in this set, they are no longer considered as candidates for recruitment.
[0041] Furthermore, the data requester notifies the selected workers to perform tasks, and each selected worker... Perform perception tasks to obtain perceived data. Use your own public key encryption Then it is uploaded to the blockchain, and the encrypted data is set as follows: The encryption process is as follows: Worker For plaintext Select the first preset random number satisfy Then, substitute the first public key. Computational Sensing Data Encryption ,in This represents the plaintext data encrypted using a homomorphic proxy re-encryption algorithm based on bilinear pairings. At this point, only the worker can access the system using their private key. generate , and then calculate Then, the calculation is performed using Pollard's Kangaroos algorithm. The discrete logarithm is used to obtain the plaintext task data. Other entities are unable to deduce the private key. The value cannot be stolen from the task data. .
[0042] Furthermore, the data disturbance server (DS, a type of server used to generate noise) uses a hash function to generate... noise value The noise value (i.e., the preset noise value) is encrypted using the first public key and sent to the blockchain. The encryption process is the same as the initial perceived data, and the resulting noise ciphertext is in the form of... The encrypted noise set is }
[0043] Smart contracts will and Performing homomorphic addition under encrypted conditions yields noisy data: .
[0044] in, This is a homomorphic addition operation.
[0045] In the above technical solution, only the sensing device can obtain the corresponding plaintext data based on the encrypted sensing data using its own private key. Other entities cannot deduce the value of the private key and therefore cannot steal the sensing data. The immutability of blockchain ensures the security and integrity of noisy data during the synthesis process, preventing data from being maliciously tampered with.
[0046] Furthermore, the step of obtaining the encrypted first noisy data synthesized on the blockchain based on the initial sensing data further includes: re-encrypting the noisy data based on the re-encryption key to obtain the encrypted first noisy data; wherein the re-encryption key is generated by the sensing device based on the second public key and the first private key.
[0047] In some embodiments, each worker uses the data requester's public key (i.e., the second public key). ) and your own private key (i.e., the first private key) Generate a re-encryption key Send it to the blockchain.
[0048] Furthermore, blockchain smart contracts utilize re-encryption keys. For noisy data After re-encryption, the generated Send to DR; the re-encryption process is as follows: according to ,in ;at this time It was then re-encrypted as .
[0049] The above technical solutions further enhance data privacy protection, making data more secure during transmission and storage.
[0050] Furthermore, obtaining the second noisy data includes: decrypting the first noisy data using a second private key to obtain the second noisy data.
[0051] In some embodiments, DR obtains the re-encrypted first noisy data from the blockchain. Based on the fundamental properties of bilinear mappings, through the second private key held... get Therefore, we can calculate: ; Then, the calculation is performed using Pollard's Kangaroos algorithm. The discrete logarithm is used to obtain the data after adding noise (i.e., the second noisy data). .
[0052] The above technical solution ensures data availability while maintaining data security by relying on the private key during the decryption process.
[0053] Step S200: Calculate the target weight based on the second noisy data, and calculate the target data true value based on the target weight and the initial sensing data.
[0054] The step of calculating the target weight based on the second noisy data includes: calculating an initial true value based on the second noisy data, obtaining an initial weight based on the initial true value and iteratively updating it; wherein, when the absolute value of the difference between the latest initial true value and the previously updated initial true value is less than a set threshold, the initial weight corresponding to the latest initial true value is taken as the target weight.
[0055] In some embodiments, based on the second noisy data Calculate the initial truth value Then repeat the following steps until convergence: Update initial weights ; Update initial truth value ; Among them, when When convergence is reached, it is determined that the convergence has occurred. The threshold value can be set by those skilled in the art, and is not limited here.
[0056] It should be noted that the final The value is the true value of the target data with noise, and the corresponding The target weight is denoted as .
[0057] In the above technical solution, the iterative update method can continuously optimize the weights, so that the target weights can more accurately reflect the true situation of the data.
[0058] Furthermore, the step of calculating the target data truth value based on the target weight and the initial sensing data includes: obtaining a first truth value based on a first public key and a random interference noise vector, and obtaining an interference noise ratio based on the first truth value and the ciphertext of the sensing data; re-encrypting the interference noise ratio based on a re-encryption key, and decrypting the re-encrypted interference noise ratio based on the re-encryption key to obtain a noisy plaintext vector; and obtaining a second truth value based on a second public key and a random interference noise vector, and decrypting the second truth value based on a second private key to obtain a target interference noise vector, while simultaneously sampling to obtain a random mask scalar.
[0059] In some embodiments, DS calculates each interference noise value. The reciprocal of the first public key is then encrypted and uploaded to the blockchain to obtain the first truth value. It should be noted that DS pre-generates a random interference noise vector. Several of these values are the interference noise values. .
[0060] Furthermore, the blockchain stores the encrypted data; through homomorphic multiplication, the encrypted interference-to-noise ratio is obtained: ; The blockchain's re-encryption key is re-encrypted and then decrypted, resulting in a plaintext vector with interference noise, i.e., a noisy plaintext vector. .
[0061] Furthermore, DS uses the second public key corresponding to DR to encrypt the interference noise value. And send it to the blockchain, at which point you will receive: ; Then DR uses the second private key to decrypt and obtain... And sample to obtain a random mask scalar. .
[0062] In the above technical solution, the noise in the data can be quantified and processed by calculating the interference-noise ratio, providing a basis for subsequent noise removal; the re-encryption process ensures the security of the interference-noise ratio during processing, and the noisy plaintext vector obtained after decryption can be used for subsequent calculations; the introduction of a random mask scalar can further protect the privacy of the data and prevent the data from being leaked during the calculation process.
[0063] Furthermore, the step of calculating the target data truth value based on the target weight and the initial perceived data further includes: obtaining a mask weighted value and an unmasked weight sum based on the target weight; encrypting the mask weighted value and the unmasked weight sum respectively based on the second public key; obtaining the mask aggregate truth value encrypted on the blockchain based on the second public key; wherein the mask aggregate truth value is obtained by decrypting the mask weighted value and the unmasked weight sum respectively on the blockchain based on the second private key and then combining it with the random mask scalar; and decrypting the encrypted mask aggregate truth value based on the second private key and removing the random mask scalar to obtain the target data truth value.
[0064] In some embodiments, based on target weights DR calculates the mask weights And calculate the unmasked weights and sums Then, it is encrypted using the second public key to obtain... And upload it to the blockchain.
[0065] Furthermore, in the context of After decryption, the blockchain assigns weights to each mask. With the corresponding ratio Multiply and sum the results: ; This is used to obtain the truth value of mask aggregation. .
[0066] Furthermore, the blockchain uses a second public key for encryption. get The data is then sent to the DR, who uses the second private key to decrypt it and remove the random mask scalar to obtain the true value of the target data. At this point, the true value of the target data is noise-free.
[0067] In the above technical solution, through a series of encryption, decryption and calculation operations, the final target data truth value can effectively remove the influence of noise and improve the data quality.
[0068] Step S300: Construct a confidence interval based on the target data truth value, verify the initial sensing data based on the confidence interval, and use the verified initial sensing data as the target sensing data.
[0069] The step of constructing a trusted interval based on the target data truth value to verify the initial sensing data includes: constructing a trusted interval based on the target data truth value and a set error; encrypting the initial sensing data using the sensing device based on a first public key to obtain homomorphic ciphertext, and constructing an affine function based on a second preset random number and a preset offset value to obtain target ciphertext and trusted endpoints based on the affine function, the homomorphic ciphertext, and the trusted interval; decrypting the target ciphertext using the sensing device based on a first private key, and comparing the decrypted target ciphertext with the trusted endpoints; wherein, based on the second preset random number, it is determined whether the target ciphertext is within the interval corresponding to the trusted endpoints; if it is, the verification is deemed successful; otherwise, the verification is deemed unsuccessful, and the corresponding initial sensing data is discarded.
[0070] In some embodiments, based on the target data truth value Construct a reliable interval with the system's preset maximum allowable error (i.e., the set error): ; in, It can be set based on historical statistics or task requirements; to simplify the expression, let , ,at this time .
[0071] Furthermore, a "claim-challenge-response-verify" protocol is adopted, in which the blockchain smart contract acts as the verifier and the worker acts as the prover. The detailed steps are as follows: Claim: Workers The initial sensing data is encrypted, and its homomorphic ciphertext is: The worker sends it to the blockchain as their claim, and the claim is subject to range verification.
[0072] Challenge: After receiving a worker's claim, the contract samples a second pre-set random number. With preset offset value Construct affine functions ;Calculation in the ciphertext domain Coordinates of the endpoints of the confidence interval and ; According to the following formula, target ciphertext The contract will , and target ciphertext It was sent to the workers as a challenge they claimed.
[0073] in, ; ; .
[0074] Response: Workers first decipher the code after receiving challenges from the contract. get Then compare , and The system calculates the size of the three values and sends the comparison result to the blockchain as a response to the contract.
[0075] Verification: The contract receives a response, and the sign of k determines whether the interval relationship is satisfied. ; If the magnitude of the worker's response satisfies the above formula, then the blockchain contract trusts the perception data provided by the worker. If the size relationship of the worker's response does not satisfy the above formula, the blockchain contract will determine that the data is an outlier. This will remove it from the encrypted set of perceived data, and the blockchain will eventually obtain a encrypted set of reliable data (i.e., the target perceived data).
[0076] In the above technical solution, the encryption process ensures the privacy of data during processing, the construction of affine functions can further transform and process the data, and data verification effectively filters out high-quality data, improving the reliability of the final target perception data.
[0077] Furthermore, it also includes updating the reputation of the sensing device based on the verification results.
[0078] In some embodiments, if If this improves the worker's credibility, then... If a worker is dishonest, their credibility will decrease; this ensures that the system gradually increases the number of trustworthy workers while suppressing dishonest workers.
[0079] Among them, let Indicates time Reputation level, initial value After the interval verification step, the binary label Indicate the validity of the reported data: ; The reputation update formula at this time is: ;in, This is a smoothing factor.
[0080] In the above technical solution, sensing devices with high reputations are more likely to be selected in subsequent data collection processes, while sensing devices with low reputations may be used less or eliminated. This incentive mechanism can encourage sensing devices to provide more accurate and reliable data, improving the performance of the entire data collection system. It should also be noted that through the dynamic reputation mechanism, the system can automatically identify and reduce the weight of low-quality or malicious data, significantly improving the accuracy of the final truth value estimation and maintaining strong robustness even when the proportion of malicious participants is high.
[0081] In summary, the data acquisition method described in this application ensures that neither the data requester nor the aggregation server can obtain the original perceived data, fundamentally preventing the leakage of participant privacy, achieving end-to-end privacy protection, and ultimately obtaining noise-free target perceived data, thus guaranteeing data quality.
[0082] Based on the same concept, please refer to Figure 2 This application also provides a data acquisition system, the system comprising: an acquisition module, configured to acquire encrypted first noisy data synthesized on a blockchain based on initial sensing data, and decrypt the first noisy data to obtain second noisy data; a calculation module, configured to calculate a target weight based on the second noisy data, and calculate a target data truth value based on the target weight and the initial sensing data; and a verification module, configured to construct a confidence interval based on the target data truth value, to verify the initial sensing data based on the confidence interval, and to use the verified initial sensing data as the target sensing data.
[0083] It should be noted that the data acquisition system and the data acquisition method are based on the same concept and have the same specific implementation method, which will not be described in detail here.
[0084] In the above technical solution, the acquisition module ensures the security and availability of the data, providing a foundation for subsequent calculations and verifications; the calculation module improves the accuracy and quality of the data and effectively removes the influence of noise; the verification module selects high-quality data as target perception data, ensuring the reliability of the data.
[0085] In addition, to verify the superiority of this application, a variety of typical algorithms were selected for comparison based on the same dataset. The evaluation was carried out from three aspects: MAE, RMSE and time cost. An ablation analysis was also performed on the VNPP algorithm to explore the contribution of its key modules to the overall performance.
[0086] like Figure 3 and Figure 4 As shown, under all malicious ratios, the method of this application significantly outperforms the baseline method in both MAE and RMSE metrics, and its advantage becomes more and more obvious as the malicious ratio increases. This significant advantage stems from the reputation management mechanism and improved truth inference strategy introduced by VNPP, which can effectively identify and weaken the impact of malicious workers during the iteration process.
[0087] Specifically, when the malicious ratio is 0.1, the final MAE of VNPP is only 0.114, which is about 86%–87% lower than that of CRH (0.846), MEAN (0.843), and MEDIAN (0.762); the RMSE is only 0.127, far lower than that of CRH (0.997) and MEAN (0.999). With a malicious ratio of 0.3–0.5, the MAE of CRH increases from 1.25 to 2.25, while VNPP remains stable at 0.190 and 0.188, with corresponding RMSEs of only 0.213 and 0.222, demonstrating excellent noise immunity. In the extreme environment of a malicious ratio of 0.7, the MAE of CRH, MEAN, and MEDIAN all exceed 4.0, and the RMSE approaches 5.0, making effective prediction difficult. In contrast, the method in this application still controls the MAE at 0.204 and maintains the RMSE at 0.241, with an error only about 1 / 20th that of CRH.
[0088] Among them, MEAN uses a simple arithmetic mean as the estimated true value; MEDIAN uses the median as the estimated true value and is more robust to outliers than MEAN; CRH is a widely used reputation-weighted truth discovery algorithm that integrates worker reports through a weighted average.
[0089] Furthermore, such as Figure 5The diagram illustrates the comparative analysis results of worker credibility (i.e., the credibility of sensing devices). Specifically, under initial conditions, the credibility of all workers except for 1% of trustworthy workers was unknown, so their credibility was set to 0.50. After 5 iterations: the credibility of malicious workers, such as worker 19, decreased from 0.50 to 0.35; the credibility of honest workers, such as worker 200 and worker 7, increased from 0.50 to over 0.60; and the credibility of wavering workers, such as worker 71 and worker 745, remained consistently within the 0.50–0.55 range. It is evident that even trustworthy workers may experience slight decreases in credibility in individual rounds due to data fluctuations, but overall maintain an upward trend. Malicious workers (such as worker 90 and worker 219) sometimes attempt to gain trust by intermittently submitting accurate data, causing a temporary increase in their credibility. However, the method described in this application effectively identifies such strategic behavior, and ultimately, the credibility of malicious workers continues to decrease.
[0090] These dynamic changes are indistinguishable by methods such as CRH, MEAN, and MEDIAN. When the method of this application is applied, the reputation gap between malicious and honest workers reaches 0.25 in the third round. This shows that the method of this application can dynamically identify worker types and achieve a balance mechanism of punishing malice, rewarding honesty, and maintaining neutrality, thereby providing more refined modeling at the worker level and enhancing long-term robustness.
[0091] Although exemplary embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above exemplary embodiments are merely illustrative and are not intended to limit the scope of this application. Various changes and modifications can be made therein by those skilled in the art without departing from the scope and spirit of this application. All such changes and modifications are intended to be included within the scope of this application as claimed in the appended claims.
[0092] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0093] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed.
[0094] The various component embodiments of this application can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some modules according to the embodiments of this application. This application can also be implemented as an apparatus program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such an implementation of this application can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.
[0095] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0096] Although the description of this application has been made in conjunction with the specific embodiments described above, it will be apparent to those skilled in the art that many substitutions, modifications, and variations can be made based on the foregoing. Therefore, all such substitutions, modifications, and variations are included within the spirit and scope of the appended claims.
Claims
1. A data acquisition method, characterized in that, Includes the following steps: Obtain the first noisy data synthesized on the blockchain based on the initial perception data, and decrypt the first noisy data to obtain the second noisy data; The target weight is calculated based on the second noisy data, and the target data true value is calculated based on the target weight and the initial perceived data. Furthermore, a confidence interval is constructed based on the target data truth value, and the initial sensing data is verified based on the confidence interval, with the verified initial sensing data being used as the target sensing data.
2. The data acquisition method according to claim 1, characterized in that, The initial sensing data originates from multiple sensing devices; the acquisition of encrypted first noisy data synthesized on the blockchain based on the initial sensing data includes: Multiple sensing devices are determined based on the credibility of the sensing devices, so as to obtain initial sensing data through the sensing devices, and to obtain the ciphertext of the sensing data by the sensing devices based on a first preset random number, a first public key and the initial sensing data; In addition, the noisy ciphertext generated based on a first preset random number, a first public key, and a preset noise value is obtained, so as to synthesize noisy data on the blockchain based on the perceived data ciphertext and the noisy ciphertext.
3. The data acquisition method according to claim 2, characterized in that, The acquisition of the encrypted first noisy data synthesized on the blockchain based on the initial perceptual data further includes: The noisy data is re-encrypted based on the re-encryption key to serve as the first noisy encrypted data; wherein the re-encryption key is generated by the sensing device based on the second public key and the first private key.
4. The data acquisition method according to claim 3, characterized in that, The process of obtaining the second noisy data includes: The first noisy data is decrypted using the second private key to obtain the second noisy data.
5. The data acquisition method according to claim 4, characterized in that, The step of calculating the target weight based on the second noisy data includes: The initial true value is calculated based on the second noisy data, and the initial weights are obtained and iteratively updated based on the initial true value; Specifically, when the absolute value of the difference between the latest initial true value and the previously updated initial true value is less than a set threshold, the initial weight corresponding to the latest initial true value is used as the target weight.
6. The data acquisition method according to claim 5, characterized in that, The step of calculating the true value of the target data based on the target weight and the initial perceived data includes: A first truth value is obtained based on the first public key and a random interference noise vector, and the interference noise ratio is obtained based on the first truth value and the ciphertext of the perceived data. The interference noise ratio is re-encrypted based on the re-encryption key, and the re-encrypted interference noise ratio is decrypted based on the re-encryption key to obtain a noisy plaintext vector. Furthermore, a second true value is obtained based on the second public key and the random interference noise vector, and the second true value is decrypted based on the second private key to obtain the target interference noise vector, while sampling to obtain a random mask scalar.
7. The data acquisition method according to claim 6, characterized in that, The step of calculating the true value of the target data based on the target weight and the initial perceived data further includes: The masked weighted value and the unmasked weight sum are obtained according to the target weight, and the masked weighted value and the unmasked weight sum are encrypted respectively based on the second public key; Obtain the truth value of the mask aggregation encrypted on the blockchain based on the second public key; wherein, the truth value of the mask aggregation is obtained by decrypting the mask weighted value and the unmasked weight sum based on the second private key on the blockchain and then combining it with the random mask scalar; Furthermore, the encrypted mask aggregate truth value is decrypted based on the second private key, and the random mask scalar is removed to obtain the target data truth value.
8. The data acquisition method according to claim 7, characterized in that, The step of constructing a confidence interval based on the target data truth value, and then verifying the initial perceived data based on the confidence interval, includes: A reliable interval is constructed based on the true value of the target data and the set error. The sensing device encrypts the initial sensing data based on the first public key to obtain homomorphic ciphertext, and constructs an affine function based on the second preset random number and preset offset value to obtain the target ciphertext and trusted endpoint based on the affine function, the homomorphic ciphertext and the trusted interval. The sensing device decrypts the target ciphertext based on the first private key, and then compares the decrypted target ciphertext with the trusted endpoint. Specifically, based on the second preset random number, it is determined whether the target ciphertext is within the interval corresponding to the trusted endpoint. If it is, the verification is deemed successful; otherwise, the verification is deemed unsuccessful, and the corresponding initial perception data is removed.
9. The data acquisition method according to claim 8, characterized in that, Also includes: The credibility of the sensing device is updated based on the verification results.
10. A system employing the data acquisition method as described in any one of claims 1-9, characterized in that, The system includes: The acquisition module is used to acquire the encrypted first noisy data synthesized on the blockchain based on the initial perception data, and to decrypt the first noisy data to obtain the second noisy data; The calculation module is used to calculate the target weight based on the second noisy data, and to calculate the target data true value based on the target weight and the initial perceived data; The system also includes a verification module, which is used to construct a confidence interval based on the target data truth value, to verify the initial sensing data based on the confidence interval, and to use the verified initial sensing data as the target sensing data.