Security-enhanced multi-party device data differential privacy protection method

Abstract

The invention provides a security-enhanced multi-party device data differential privacy protection method, which comprises the following steps that: a user acquires data from a device sensor to form multi-user data, and uploads the multi-user data to a block chain network system; when a querier initiates a data query request to the block chain network system, the block chain network system judges whether the query request is single user data query or all user data query; if the query is single user data query, adding differential privacy noise to the user data; if all user data query is carried out, differential privacy noise is added to all user data, and noise cancellation can be carried out on the differential privacy noise of all user data, so that data and privacy of individual users are hidden in all data.

Description

Technical Field

[0001] The present invention relates to a security-enhanced multi-party device data differential privacy protection scheme, belonging to the field of blockchain technology and privacy security. Background Art

[0002] The development of IIOT has brought tremendous changes and opportunities to the industrial field. It has greatly promoted data sharing, intelligent decision-making and cross-enterprise collaboration among multiple enterprises. Under the "device data feedback" model of IIOT, enterprises purchase equipment from manufacturers and regularly provide daily operation data of equipment to manufacturers according to the contribution agreement to support manufacturers in equipment maintenance and iteration; provide enterprises with better new generation products and technical support to improve the production efficiency of enterprises. The startup, collection and processing of large amounts of data by the sensor equipment at the physical layer, and the transmission to the IIOT network using the original and centralized processing method of the third-party platform, pose privacy risks to all parties. Now blockchain, as a representative of transparent and open distributed storage systems, is widely used in data sharing. The existing blockchain system architecture splits the upload and access blockchain functions, sets them locally, and then places the consensus nodes in the cloud, which protects the security of data to a certain extent, but there is still no way to solve attacks based on background knowledge.

[0003] Differential privacy technology can solve the above problems by adding standardized random noise to the data to confuse the real value. However, differential privacy algorithms are often used as third-party services or privacy tools, placed outside the blockchain system and in parallel with the blockchain system. In this way, the third-party server will inevitably involve the original data. In addition, the statistical data of enterprise equipment consists of multiple attribute parameters, such as equipment wear, heat dissipation rate, power failure, etc. For attackers, even if the noise changes in the statistical values ​​can cause some confusion to them, the trend formed by multiple values ​​will still expose some privacy information. Therefore, the privacy issues existing in manufacturer-enterprise equipment data sharing have not been effectively solved. Summary of the invention

[0004] In order to solve the problems existing in the prior art, the present invention provides a security-enhanced multi-party device data differential privacy protection method that makes different response results according to different query purposes of data collectors.

[0005] In order to achieve the above object, the technical solution proposed by the present invention is: a security-enhanced multi-party device data differential privacy protection method, comprising:

[0006] Users collect data from device sensors to form multi-user data, and upload the multi-user data to the blockchain network system;

[0007] When the queryer initiates a data query request to the blockchain network system, the blockchain network system determines whether the query request is a single user data query or a full user data query;

[0008] If it is a single user data query, differential privacy noise is added to the user data; if it is a query of all user data, differential privacy noise is added to all user data, and the differential privacy noise of all user data can be noise-cancelled.

[0009] A further design of the above technical solution is: adding local differential privacy noise to the multi-user data before uploading the multi-user data to the blockchain network system.

[0010] The local differential privacy is a noise mechanism based on Gaussian distribution. The noise distribution formula is:

[0011] in, where s is the sensitivity of f, represents the mean as 0, ε as the privacy budget, and the variance as σ 2 Gaussian distribution sampling results.

[0012] The blockchain network system is provided with a smart contract module, which is provided with a storage module, a differential privacy module and a query module; the storage module is used to store multi-user data, the query module is used to judge the query request, and the differential privacy module is used to add differential privacy noise to the multi-user data.

[0013] Users upload multi-user data to the blockchain network system through the client; inquirers initiate data query requests to the blockchain network system through the client.

[0014] The blockchain network system needs to be initialized before uploading multi-user data to the blockchain network system.

[0015] Multi-user data is data that includes multiple parameters of a device.

[0016] The blockchain network system can offset the differential privacy noise added to the same type of data in all user data.

[0017] Compared with the prior art, the present invention has the following advantages:

[0018] After the multi-user data of the device is uploaded to the blockchain, the present invention saves the data in the storage module of the smart contract module, and the data is read inside the smart contract, which reduces the steps of data processing and the risk of privacy leakage; the differential privacy module is also integrated in the smart contract module, without the need for a third-party server, and can better protect data security. In addition, in the specific scenario of device data feedback from manufacturers and multiple enterprises, in order to target the manufacturer's potential privacy snooping behavior, if a specific query is made on a certain enterprise, the smart contract adds enough noise to distort the various data of the enterprise and there is no noise offset by other enterprise data, making the data of individual enterprises unavailable; when the manufacturer queries all enterprises, the smart contract adds enough noise to distort the various data of the enterprise and the various data of other enterprises can be noise offset, making the data and privacy of individual enterprises hidden in the overall data. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a flowchart of the privacy protection service of the present invention.

[0020] Figure 2 This is the architecture diagram of the privacy protection solution of the present invention.

[0021] Figure 3 This is a diagram illustrating the privacy protection of data upload and query in the present invention.

[0022] Figure 4 This is a comparison chart of the single query response and query protection of the present invention.

[0023] Figure 5 This is a comparison chart of privacy protection for all query responses of the present invention. DETAILED DESCRIPTION

[0024] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0025] Embodiment 1

[0026] like Figure 1 and Figure 2 As shown, the security-enhanced multi-party device data differential privacy protection method of this embodiment includes the following steps:

[0027] Step 1: Deploy the blockchain network system and complete initialization; specifically including:

[0028] Step 1.1: The blockchain generates encryption materials, including identity certificates and private keys, and generates node and channel configuration files, and performs the first data synchronization and consensus operation with all client nodes.

[0029] Step 1.2: Start the Orderer node as the initial node of the blockchain network. Create a channel configuration file to create a new channel, and join the Peer node to the channel. Install and instantiate the chain code.

[0030] Step 1.3: After the network nodes complete the first consensus, the blockchain system is successfully started.

[0031] Step 2: Users collect operation and maintenance data from device sensors and aggregate the data into statistical data. To protect their privacy, users add local differential privacy noise to the statistical data, such as Figure 2 ① step; specifically including:

[0032] Step 2.1: The user collects data from local device sensors, processes the original device fault data into statistical data of each fault cause parameter, and aggregates the data to form multi-user data.

[0033] Step 2.2: In order to protect their own privacy, the user adds localized differential privacy noise to the statistical data collected from the device. The noise used is privacy noise based on the Gaussian mechanism. The noise distribution formula is: in where s is the sensitivity of f, represents the mean as 0, ε as the privacy budget, and the variance as σ 2 The Gaussian (normal) distribution sampling result. Its characteristic is that the perturbed data increases the availability of the perturbed data with a greater probability p near the real data, and the perturbed data with a smaller probability q to protect the privacy of the real data. The parameter that controls the range of data perturbation is b, and the value of b decreases as ε increases, that is, the privacy protection strength decreases. The transaction can be expressed as U = {f1, f2, f3…f N}, where N represents N enterprises, transactions that have been protected by local privacy protection

[0034] Step 2.3: In this embodiment, local noise is added to the uploaded data, as shown in Algorithm 1 below.

[0035]

[0036] Step 3: The user uploads the multi-user data to the blockchain network system through the client; specifically including:

[0037] Step 3.1: Blockchain transaction generation and endorsement: Whether querying data or uploading data, it can be regarded as a transaction. First, the client submits a transaction request to the designated endorsement node, which includes the chain code to be executed and the corresponding parameters. This process does not change the state of the ledger. Members of the same channel will be required to perform endorsement. Then the module execution result is sent to other endorsement nodes that need to endorse the transaction.

[0038] Step 3.2: Blockchain node sorting and broadcast verification: The sorting node receives enough endorsement results from the endorsing nodes and sorts them. The sorting node packages the sorted endorsement results into a block and broadcasts it to all verification nodes. The verification node receives the block and verifies the transactions contained in the block. If the verification passes, the verification node adds the block to the local blockchain and updates the status ledger record, such as Figure 2 Finally, the submission success notification is returned to the client.

[0039] Step 4: The manufacturer, i.e. the inquirer, initiates a data query request to the blockchain; a range query is performed by specifying two parameters: user name and upload time period date.

[0040] Step 5: The smart contract module of the blockchain determines whether the query request is about a single user data query or a query of all user data, and adds corresponding differential privacy noise based on the judgment result; specifically, it includes:

[0041] Step 5.1: To address the privacy issues in query transactions during device data feedback sharing, this embodiment integrates the differential privacy mechanism into the chain code, such as Figure 3 , analyze the user parameters of the query request. Assume that the total number of enterprise users cooperating with the manufacturer is N, and the user parameter of the manufacturer's query f is n. The query transaction is represented by Q = {q1, q2, q3...q K}, the number of enterprises involved in each query q is n, and the transaction response is expressed as Each query response from the manufacturer is actually a statistics of n users' data.

[0042] Step 5.2: For the data of N users associated with the manufacturer, when querying, the blockchain calls the smart contract to add a large Laplace noise (denoted as noise) to each user's histogram data. The relationship is as follows:

[0043] q=f * +noise (1)

[0044] Step 5.3: Thoroughly interfere with the numerical values of the histogram data and the trends of each attribute parameter. Therefore, when n ≤ N, the manufacturer's query response will be distorted, thus avoiding intentional and linked queries for individual users. In addition, the smart contract sets the sum of all noises of N users to be 0. That is, when n = N, the manufacturer's query is non-deliberate, and the chain code will sum up all of them, which will cancel each other out. The relationship is as follows. Algorithm 2 is the query noise addition algorithm.

[0045]

[0046]

[0047]

[0048]

[0049] Step 5.4: The algorithm determines the relationship between n and N in lines 2 and 7 to determine whether the manufacturer's query this time may be a link attack. When n < N, because there is no cancellation of multiple noises, a relatively large noise will be added to the data. The algorithm idea aims to not expose individual user data alone, but to hide it among the data of multiple users, such as Figure 2 in step ②.

[0050] In this embodiment, for the range query of the name and date of the data collector, the smart contract adds enough distorting noise to various types of data of each user related to the data collector, and the noises of the same type of data of each user can cancel each other out.

[0051] Specifically, when the data collector conducts a targeted query on a certain user, the smart contract adds enough distorting noise to various types of data of the user and there is no noise cancellation by other user data; when the data collector conducts a query on all users, the smart contract adds enough distorting noise to various types of data of the users and the various types of data of other users can cancel each other out.

[0052] Step 6: The differential privacy module of the smart contract aggregates and publishes the processed data, and feeds back the query results to the manufacturer in the form of a histogram through the client web page, such as Figure 2 in step ④.

[0053] This embodiment integrates the above into the chain code (smart contract) of fabric. The system model consists of a client program, an endorsement-ordering node, and a privacy protection module. The detailed information of each functional module is as follows:

[0054] Original Equipment Manufacturer (OEM): Manufacturers obtain data stored in the blockchain ledger by the enterprise through client queries. Managers analyze the proportion trend information of the data to determine the optimization focus and direction of future products. In order to obtain data more accurately, manufacturers may specify multiple queries for histogram data of one or more enterprises in different time periods. In addition, manufacturers and enterprises reach a reward agreement on data provision in advance, which determines the accuracy of the data provided by the enterprise.

[0055] Enterprise users: Enterprises will reach an agreement with manufacturers in advance on providing equipment data. Enterprises will collect equipment operation status data from purchased equipment and summarize the operation data into statistical data. In order to protect their own privacy, enterprises can add differential privacy noise by themselves when uploading data or use smart contracts as agents. In addition, the standard for adding noise is determined by the reward agreement.

[0056] Client program: The client program is the component responsible for sending request transactions between blockchain nodes and blockchain networks. The client program provides services such as writing to the ledger, searching the ledger, querying the ledger, and updating the ledger to the nodes. Manufacturers and partner companies use the client to query or submit data.

[0057] Endorsement and sorting nodes: In the fabric network, transactions are first sent to the endorsing nodes before being submitted to the blockchain network. The endorsing nodes execute and verify the transactions. If the verification is successful, it will generate an endorsement signature. The client sends a transaction proposal with an endorsement signature to the sorting node. After the sorting node collects transactions in the network and reaches a consensus, it adds the transaction to the blockchain.

[0058] Privacy protection module: The above method integrates differential privacy into the blockchain chain code, and uses the transparency and tamper-proofness of the blockchain to enhance the privacy protection and fairness of the scheme. The present invention proposes two functions to handle the privacy issues that arise when querying and submitting data respectively.

[0059] The transactions in the system model of this embodiment are divided into query transactions and update transactions. In addition, query transactions refer to the union of read transactions and range transactions. For example, an OEM wants to query the submitted data of several enterprises in a certain period of time. The query transaction is represented by Q = {q1, q2, q3...q K}, the number of enterprises involved in each query q is n, and the transaction response is expressed as

[0060] In the threat model of the present invention, there are many privacy concerns about the security of individual user data. First, one or more manufacturers, which show a strong interest in the potential privacy of multiple enterprises and peers (regarded as potential opponents) with which they cooperate, send query requests to spy on the business secrets of their partners. Because the data is encrypted or noise is added, these query responses may be considered reasonable or privacy will not be pried out. In fact, these protection measures can be bypassed by some other special methods, such as link attacks and combination attacks. Suppose that an OEM is a curious person who is interested in understanding the real equipment data of one of the enterprise partners to predict its production capacity and benefits; he can combine and compare the query results with the known background knowledge (information of other partners) to obtain the expected results. In the second case, the curious person initiates a sharing request to members of the same channel. Since the members share the same account book, even if the response is optimized, the curious person can still reveal the actual results through professional means such as matching and comparison.

[0061] In addition, the curious person considered by the present invention cannot edit or delete data, but can share data with others. By purposefully and specifically repeating the request for data, and then processing the response results (for example, taking the average), the actual results can be obtained even if the data with differential privacy noise is added. In this scenario, a single enterprise data consists of statistical histograms of multiple attributes. The curious person wants to obtain device status data, especially the distribution trend in the data. Compared with the trend, the subtle changes in the numerical value are understandable to the curious person, which is different from other protection scenarios of differential privacy. Therefore, the effect of a single perturbation of the numerical value is limited, and how to protect the trend of independent histograms is the focus of the work.

[0062] Based on the above threats, this embodiment proposes the above method. When the queryer queries a single user A, the smart contract can detect the query transaction and judge the curiosity of the queryer. A's data will be added with a second noise. This noise can destroy the original trend of the statistical data. Figure 4 As shown in the figure, the same device data exhibits very different trends in the upload phase and the query phase.

[0063] At the same time Figure 5 As shown in the figure, when the queryer conducts multi-user queries on other cooperating users, the noise added by the smart contract will almost cancel out in the multi-user query. Obviously, the queryer gets a more accurate feedback in the second query, but this is the beginning of his doubts. Because of his knowledge background and the two very different query results, he is confused and meaningless about the inferred information of A. The inference paradox is created by using the queryer's knowledge background and the large difference in query results, thereby protecting the user's data privacy.

[0064] Embodiment 2

[0065] A security-enhanced multi-party device data differential privacy protection system of this embodiment includes several local client interfaces, a customer service application module and a blockchain network set in the cloud, and several data processing modules in the network. The user maintains statistics on the raw data collected from the device sensor, adds localized differential privacy noise to the raw statistical data, and then uploads it to the blockchain through the local client interface. The blockchain uses the storage module to consensus, sort, and store the data. The data collector also requests access to the data through the client web interface. The blockchain smart contract receives and retrieves the data according to the parameters name and date of the access request. The data is processed by the differential privacy processing module integrated in the chain code and returned to the web page, and displayed in the form of a bar chart.

[0066] The above description is only a preferred embodiment of the present invention, and therefore cannot be used to limit the scope of implementation of the present invention. That is, equivalent changes and modifications made according to the scope of the patent application of the present invention and the contents of the specification should still fall within the scope covered by the patent of the present invention.

Claims

1. A security-enhanced multi-party device data differential privacy protection method, characterized by: Users collect data from device sensors to form multi-user data, and upload the multi-user data to the blockchain network system; When the queryer initiates a data query request to the blockchain network system, the blockchain network system determines whether the query request is a single user data query or a full user data query; If it is a single user data query, differential privacy noise is added to the user data; if it is a query of all user data, differential privacy noise is added to all user data, and the differential privacy noise of all user data can be noise-cancelled.

2. The method for protecting differential privacy of multi-party device data with enhanced security according to claim 1, characterized in that: Local differential privacy noise is added to multi-user data before uploading it to the blockchain network system.

3. The method for protecting differential privacy of multi-party device data with enhanced security according to claim 2, characterized in that: The local differential privacy is a noise mechanism based on Gaussian distribution. The noise distribution formula is: in, where s is the sensitivity of f, represents the mean as 0, ε as the privacy budget, and the variance as σ 2 Gaussian distribution sampling results.

4. The method for protecting differential privacy of multi-party device data with enhanced security according to claim 1, characterized in that: The blockchain network system is provided with a smart contract module, which is provided with a storage module, a differential privacy module and a query module; the storage module is used to store multi-user data, the query module is used to judge the query request, and the differential privacy module is used to add differential privacy noise to the multi-user data.

5. The method for protecting differential privacy of multi-party device data with enhanced security according to claim 4, characterized in that: Users upload multi-user data to the blockchain network system through the client; inquirers initiate data query requests to the blockchain network system through the client.

6. The method for protecting differential privacy of multi-party device data with enhanced security according to claim 2, characterized in that: The blockchain network system needs to be initialized before uploading multi-user data to the blockchain network system.

7. The method for protecting differential privacy of multi-party device data with enhanced security according to claim 1, characterized in that: Multi-user data is data that includes multiple parameters of a device.

8. The method for protecting differential privacy of multi-party device data with enhanced security according to claim 7, characterized in that: The blockchain network system can offset the differential privacy noise added to the same type of data in all user data.

Images