Blockchain-based Manufacturing Industry Privacy Data Security Protection System and Method

Abstract

The present invention provides a blockchain-based manufacturing industry privacy data security protection system and method, which relates to the field of network security technology. The present invention proposes a data offloading architecture based on blockchain security verification, and designs a smart contract inside the blockchain, which can conduct security supervision, integrity supervision and anti-tampering supervision on user data. The blockchain technology is used to supervise the offloading of industrial data, and a smart contract is designed to audit data during the offloading process to ensure the privacy and security of enterprises. In addition, an industrial data offloading scheme based on Asynchronous Advantage Actor-critic is used to solve the problems of delay and energy consumption cost brought by the combination of blockchain and intelligent industry.

Description

Technical Field

[0001] The present invention relates to the field of network security technology, and in particular to a blockchain-based privacy data security protection system and method for the manufacturing industry. Background Art

[0002] In the industry, privacy data such as transaction data and enterprise information are protected by both morality and regulations. There have been some architecture designs for data offloading and strict access control after data offloading. However, these studies do not consider the problem of illegal data offloading on mobile edge computing servers. In addition, unauthorized malicious Internet of Things devices may offload data to mobile edge computing servers, thereby obtaining computing resources from the edge servers. For example, an attacker controls a large number of mobile devices at multiple locations and attacks the edge mobile edge computing s at the same time, which may lead to privacy problems in data offloading. In this case, in order to ensure security protection in industrial data offloading, blockchain, with its unique decentralized, traceable, and immutable features, has become a disruptive technology by utilizing data ledgers and smart contracts. Therefore, the computing efficiency of mobile edge computing and the security features of blockchain are very valuable for intelligent industrial services in a distributed industrial environment. So far, some blockchain-based intelligent industrial solutions have been designed to enhance architecture security. However, existing research mainly focuses on aspects such as industrial data security sharing, access control, and auditing after data offloading, and ignores the security protection during the industrial privacy data offloading process.

[0003] In addition, using blockchain technology to offload industrial data will also generate latency costs and energy consumption costs. Existing research ignores these costs when considering using blockchain technology to solve privacy problems in the industrial data offloading process.

[0004] In the current prior art, CN114547210B, a medical data security sharing system based on a dual blockchain, provides fine-grained data access while ensuring the secure and efficient sharing of data. CN114466409B, a control method and device for data offloading for machine communication, proposes a distributed system architecture based on federated edge learning for data offloading, optimizes the transmission power, computing power, and data offloading volume of clients, and obtains the minimum data offloading latency. However, the methods in the prior art do not consider the cost and data offloading security issues brought by blockchain technology. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention provides a blockchain-based privacy data security protection system and method for the manufacturing industry, which uses blockchain technology to supervise the offloading of industrial data for the privacy and security protection of industrial data edge computing.

[0006] On the one hand, a blockchain-based privacy data security protection system for the manufacturing industry specifically includes a data collection module, a data offloading decision module, a data local processing module, a blockchain smart contract auditing module, and a data edge processing module.

[0007] The data collection module is used to collect data of each enterprise, that is, task processing information.

[0008] The data offloading decision module receives task processing information, formulates the cost problem as a Markov decision process, and solves the data offloading decision problem through a deep reinforcement learning method based on policy search, specifically including load distribution decision, allocation of computing resources and radio transmission bandwidth, and blockchain data security auditing.

[0009] For the information that the data local processing module finally decides to process locally after the data offloading decision module, it calls the local server for processing.

[0010] For the information that the blockchain smart contract auditing module finally decides to process at the edge after the offloading decision module, it counts the data before the data is transmitted to the edge side, and conducts smart contract auditing on the data when it reaches the edge side to check that the data has no modification, replacement, loss, or addition.

[0011] For the information that the data edge processing module finally decides to process at the edge after the data offloading decision module, it calls the server at the edge side for processing.

[0012] On the other hand, a blockchain-based privacy data security protection system for the manufacturing industry is implemented based on the aforementioned blockchain-based privacy data security protection system for the manufacturing industry, and includes the following steps:

[0013] Step 1: Collect privacy data of the manufacturing industry.

[0014] The mobile device is used to collect industrial data from intelligent devices. After obtaining permission, the local edge service provider, on the one hand, allows the mobile device to offload data locally on the blockchain network; on the other hand, it packages the data collected from the manufacturing industry and participates in global blockchain data sharing.

[0015] Step 2: Make an offloading decision for the processing task of privacy data, and through blockchain smart contract technology, audit the data to ensure the security of the data offloaded to the edge side.

[0016] Suppose there are N randomly distributed mobile users Each user n Generates a computing task Tn = {Dn, Xn, Cn}, where D n Represents the total data volume of the computing task T of user n n xn represents the proportion of the computing tasks of mobile user n offloaded to the mobile edge computing of the edge server, and Cn represents the number of mobile CPU cycles consumed by user n to process unit data;

[0017] Step 2.1: Latency of data offloading

[0018] The latency includes: the latency of the data offloading edge computing part and the latency T' of the blockchain verification part during the offloading process n ; The latency of the data offloading edge computing part includes: the execution time of offloading to the local the processing time of offloading to the mobile edge computing server Among them includes three parts: the task encryption time at the local end the time of sending the task to the mobile edge computing server and the task execution time on the mobile edge computing server

[0019] Among them, D n is the size / bit of the input data, are the CPU utilization rates of the mobile / edge devices respectively, cycles / bit, refer to the CPU cycle frequencies of the local / edge devices respectively, cycles / s, x n refers to the proportion of data offloaded to the edge server, refers to the CPU utilization rate of the local device for encryption processing:

[0020]

[0021]

[0022] Among them, w n is the proportion of the bandwidth resource allocated to the nth task, B is the total radio bandwidth, p n is the transmission power per unit time of mobile user n, h n is the channel gain between the mobile device and the edge computing server, and N0 is the Gaussian white noise. R n is the uplink transmission rate, with the unit of bit / s and is defined as:

[0023]

[0024] And the latency T' of the blockchain verification part during the offloading process n respectively includes: 1) the time for the blockchain manager to send the unverified block 2) the time for block verification 3) Broadcast and comparison of verification results between verifiers ψ(n′*B)m′; 4) Time to send verification feedback

[0025]

[0026] where m′ is the number of selected verifiers, with the maximum and minimum values being v and M respectively, n′ is the number of transactions per block, with the maximum and minimum values being x and t respectively, B is the transaction size, K is the computing resource required for the block verification task, x i is the computing resource available at verifier i, O is the size of the verification feedback, R n and r u are the downlink and uplink transmission rates from the blockchain manager to the verifier respectively. ψ is a predefined parameter obtained using the statistics of block verification;

[0027] Since the task T of user n n can be offloaded to local and mobile edge servers, the total latency for simultaneously processing data at local and mobile edge servers to complete the entire offloading task is:

[0028]

[0029] Step 2.2: Energy consumption of data offloading;

[0030] The energy consumption E n includes the energy consumption of the data offloading edge computing part and the energy consumption of the blockchain verification part during the offloading process The energy consumption of the data offloading edge computing part includes local energy consumption the energy consumption of offloading to mobile edge computing for processing which includes encryption energy consumption transmission energy consumption and computing energy consumption

[0031]

[0032]

[0033]

[0034]

[0035]

[0036]

[0037] where, P M , P sec, P T , P C are the energies consumed per unit time for local data processing, local data encryption, data transmission, and edge data computing respectively.

[0038] In the blockchain, all transactions have a cost measured in gas units, which is regarded as a measure of the standardized contract cost. The energy consumption of the blockchain verification part during the offloading process Among them, is the maximum amount of gas for executing a transaction, ξ is the natural gas price value, the ratio paid to the miner per unit of natural gas; g n is the amount of gas used when executing the access control contract, and Ek refers to the energy consumed to generate a unit of natural gas:

[0039]

[0040] Step 2.3 Normalization processing:

[0041]

[0042] where α and β are the proportions of each part respectively, and α + β = 1, are the respective parts, the maximum values of latency and energy.

[0043] Step 2.4 Formulate the objective function:

[0044] Formulate the multi-user secure offloading joint optimization offloading decision A, blockchain configuration B = {m′, n′}, computing resource allocation f = [f1, f2,..., f N , and the allocation of communication resource radio bandwidth w = [w1, w2,..., w N , and the use of smart contracts:

[0045]

[0046] Constraints:

[0047] (C1) x n ∈ [0, 1] n ∈ N′,

[0048] (C2) v ≤ m′ ≤ M,

[0049] (C3) t ≤ n′ ≤ x,

[0050] (C4)

[0051] (C5)

[0052] (C6) 0 < w n ≤ 1, n ∈ N′,

[0053] (C7)

[0054] (C8)

[0055] Among them, (C1) - specific offloading allocation constraint; C(2), C(3) - blockchain configuration constraints; C(4), C(5) - mobile edge computing computing power constraints, computing resource constraints; C(6), C(7) - bandwidth allocation constraints; C(8) - smart contract constraints.

[0056] Step 2.5: For the data offloading problem, a deep reinforcement learning offloading ADRLO scheme is designed;

[0057] Step 2.5.1: Data security supervision process on the blockchain based on smart contract technology

[0058] For the enterprise's data, the enterprise first uses the Hash function to generate a hash table of the data, and uses its own private key to encrypt the hash table to generate a digital signature, and then uploads the data. The digital signature and the encrypted data are sent to the industrial data service provider together; after receiving the data, the industrial data service provider decrypts the digital signature using the enterprise's public key. According to this method, the hash table of the encrypted data is obtained; the industrial data service provider uses the Hash function on the data and compares the obtained data with the digital signature in the previous step; if the two are consistent, it means that the data has not been modified and the privacy of the encrypted data has not been damaged.

[0059] Step 2.5.1.1: The enterprise registers, and the privacy data security protection system designs public and private keys for the enterprise; the public and private keys are {a, b, u, f, spk, v} and {ω, q, ssk} respectively, where ω, q are random numbers, spk and ssk are randomly generated key pairs, v ∈ L1, u, f ∈ L2, and L1 and L2 are multiplicative cyclic groups. v, u, f are the generators of the multiplicative cyclic group, and the random numbers a, b satisfy the following equations:

[0060] v = (ω + q) 2 (a + b) ω+q

[0061] Step 2.5.1.2: According to the offloading decision, the enterprise divides the data offloaded to the mobile edge computing server into J blocks, denoted as the set S, S = {m i}, i ∈ {1, 2,..., J}, where i refers to the subscript of the J data blocks. m i represents the i-th data block. The enterprise uses the private key {ω, q, ssk} to calculate the signatures of the J blocks of data, and represents R(m i ) as the index of the corresponding hash table after hashing, then the root of the hash tree is RMHT , obtain the processed data {S, Φ}, where the signature set Φ is defined as:

[0062]

[0063] where σ i represents the signature of data block m i .

[0064] Step 2.5.1.3: Send the processed data {S, Φ} to the industrial data service provider for authorized data storage;

[0065] Industrial data service provider processing: After the industrial data service provider receives the processed data, if the latest nonce value used to generate the random number is directly obtained from Ethereum; it is not secure because a malicious attacker controls the generation of the random number. At this time, a delay function is adopted to prevent the service provider from controlling the generation of the random number in the case where the industrial data service provider is untrusted; specifically, execute the delay function f time , obtain the seed seed for generating the random number, specifically as follows:

[0066] seed = f time (nonce new )

[0067] where time is the delay time and nonce new is a new nonce value; after executing the delay function, the smart contract will verify the time value of point J, and the specific data is the set {(i, f time (nonce new ))}, i ∈ {1, 2,..., J}; when the verification is successful, a challenge, that is, verification information, will be created; if the verification fails, this service will be rejected and a suitable seed set will be selected again.

[0068] The smart contract is as follows: When the system receives the data provided by the industrial data service provider, it will check its identity to determine whether its address is the same as the previously registered address; if the audit is successful, the corresponding challenge chal = {(i, v i )}, i ∈ {1, 2,..., J}, and send it to the industrial data service provider, where v i represents the values of the randomly selected J seeds seed. If the audit is unsuccessful, the industrial data service provider will be punished.

[0069] Step 2.5.1.4: After the industrial data service provider receives the challenge sent by the smart contract, calculate the proof;

[0070] Step 2.5.1.4.1: After receiving the challenge chal, the edge computing service provider selects random numbers l, h ∈ Zp , where Z p is the set of non - negative integers less than the constant p.

[0071] Step 2.5.1.4.2: The industrial data service provider calculates the proof Proof = {Rand, μ, σ, ζ} according to the following formula, where Rand, μ, σ are the proof sub - information constructed for this proof

[0072] Rand = [(u + f) ω+q l , Rand ∈ L1

[0073]

[0074]

[0075]

[0076] Step 2.5.1.4.3: Send the proof Proof to the smart contract for verification.

[0077] Step 2.5.1.5: The smart contract verifies the proof generated by the industrial data service provider;

[0078] Create an object EnterprisePe = EnterpriseAddr(Addr) according to the enterprise address, initialize the variable sum = 0, traverse index from 0 to Pe.ChalNum.i.length, and calculate sum: sum = sum + Pe.ChalNum.i[index] * Pe.chal.v i [index]. Calculate and judge whether e(σ * Rand h*Rand , a + b) is equal to , where e(*, *) refers to the bilinear mapping; if they are equal, the verification result is successful and the data is allowed to be processed at the edge side, otherwise, the edge service provider is punished.

[0079] Step 2.5.2: The data calculation offloading algorithm based on deep reinforcement learning DRL

[0080] Propose the state space, action space, and reward function in the deep reinforcement learning offloading ADRLO, which are defined as a Markov decision process as follows:

[0081] 1) State space: The state space describes the amount of remaining resources:

[0082] ·ec: The computing resources available on the mobile edge computing server

[0083] ​· bw: Bandwidth resources available for industrial mobile edge computing servers

[0084] · ag: Maximum gas limit for security supervision provided by blockchain audit contracts

[0085] Therefore, the state of the architecture is represented as:

[0086] State = [ec, bw, ag]

[0087] 2) Action space: Define the action space as four vector sets, namely: offloading decision vector X, mobile edge computing server resource allocation vector G, radio resource bandwidth allocation vector w, and smart contract gas allocation vector set g:

[0088] · X = [X1,..., X n ,..., X N , where X n is the proportion of computing tasks allocated by enterprise n to the mobile edge computing server.

[0089] · G = [G1,..., Gn,..., G N , where G n is the size of the computing resources of mobile edge computing s for the tasks of enterprise n.

[0090] · w = [w1,..., w n ,..., w N , where w n represents the ratio of the wireless transmission rate allocated to enterprise n.

[0091] · g = [g1,..., g n ,..., g N , where g n is the amount of gas consumed for the security supervision of the audit contract of enterprise n.

[0092] Therefore, the action of the system is represented as:

[0093] Action = [X, G, w, g] = [X1, G1, w1, g1,..., X N , G N , w N , g N

[0094] ​3) Reward: The task of the agent is to continuously explore new actions in the environment to obtain the maximum reward. The agent finds the optimal offloading operation for the industrial data in each state. The goal is to minimize the total cost in the architecture, i.e., the task computing cost and the blockchain verification cost. Minimize the total cost C(s, a) of the latency cost T(s, a) and the energy cost E(s, a), and set the reward as:

[0095] r(s, a) = -C(s, a) = -(T(s, a) + E(s, a)) = -L

[0096] 4) ADRLO algorithm: The policy function π θ (a|s), that is, the actor will learn the policy to obtain the highest reward. The value function V π (s) is used to implement the single-step update of the algorithm. In the policy gradient, we update the parameter θ and calculate the gradient as follows:

[0097]

[0098] where H represents the number of τ and the random number τ ∈ [0, 1], b′ is the baseline, and T′ n represents the time period. is to discount the future reward, and γ t′-t is the discount factor. refers to the future reward. refers to the update of the action control parameter θ of the agent, using the expected value of the cumulative reward to replace the reward from t to T′ n time period: refers to the Q function, i.e., the evaluation network. E[*] is the expected value, and use the value function to replace the baseline b. does not involve actions. involves actions. is the expected value of. Therefore, and are positive or negative numbers, and the Actor - critic calculates the gradient as follows.

[0099]

[0100] Use the value of the V network to replace the value of the Q network, obtain the reward r, and jump to the state r by taking the action a in the state s t+1 . The expected value of is equal to the Q function:

[0101]

[0102] Getting closer to the expected value, replace the Q function with i.e., the current reward plus the value function of the next state In the ADRLO environment, a participant interacts with the environment through policy π to generate data. After collecting the data in the policy gradient method, the policy is updated by first estimating the value function with this data and using temporal difference or Monte Carlo to estimate the value function, and updating policy π according to the value function. The ADRLO calculates the gradient as follows:

[0103]

[0104] After obtaining π, interact with the environment to collect new data, estimate the value function, and use the new value function to update the policy and actor.

[0105] The beneficial effects of adopting the above technical solutions are as follows:

[0106] The present invention provides a blockchain-based manufacturing industry privacy data security protection system and method, which uses deep reinforcement learning technology to guide data offloading and has the following advantages: (1) jointly considering offloading decisions, computing resources, radio transmission bandwidth allocation, and blockchain data auditing, providing ideas for the resource scheduling optimization problem under the industrial blockchain architecture; (2) having significant advantages in reducing latency and energy consumption, indicating the feasibility of the solution of the present invention in industrial data offloading applications.

[0107] In addition, the present invention can effectively solve the data offloading security problem. On this basis, the present invention automatically audits the transmitted data, supervises the integrity of user data, and anti-tampering supervision by using smart contract technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0108] Figure 1 It is a flowchart of the privacy data security protection method in an embodiment of the present invention.

[0109] Figure 2 It is a schematic diagram of the intelligent model combining blockchain smart contract technology and deep reinforcement learning data offloading of the present invention

[0110] Figure 3 It is a comparison graph of relevant times of the audit contract for evaluating the experiment and the number of passes for detecting attacks.

[0111] Figure 4 It is a comparison graph of latency, energy consumption, and total cost of various algorithms in an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0112] The specific embodiments of the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. The following embodiments are used to illustrate the present invention, but are not used to limit the scope of the present invention.

[0113] On the one hand, a privacy data security protection system for the manufacturing industry based on blockchain, as Figure 2 shown, specifically includes: a data collection module, a data offloading decision module, a data local processing module, a blockchain smart contract auditing module, and a data edge processing module.

[0114] The data collection module: includes many information sensing devices of the Internet of Things, such as temperature sensors, location tracking devices, and heart rate monitoring wearable devices, etc. It is used to collect data of each enterprise, that is, task processing information;

[0115] The data offloading decision module receives task processing information, formulates the cost problem as a Markov decision process, and solves the data offloading decision problem through a deep reinforcement learning (Asynchronous Advantage Actor-Critic) method based on policy search, specifically including load distribution decision, allocation of computing resources and radio transmission bandwidth, and blockchain data security auditing, and reasonably takes measures to process data processing tasks.

[0116] The data local processing module processes the information that the data offloading decision module finally decides to process locally by calling the local server.

[0117] The blockchain smart contract auditing module counts the data before the data is transmitted to the edge end for the information that the offloading decision module finally decides to process at the edge, and conducts smart contract auditing on the data when it reaches the edge end. To check that the data has not been modified or replaced, and there is no loss or addition.

[0118] The data edge processing module processes the information that the data offloading decision module finally decides to process at the edge by calling the server at the edge end.

[0119] The processing tasks of the data local processing module do not require security auditing, while the processing tasks of the data edge processing module require auditing by the blockchain smart contract auditing contract module. After auditing, the task processing can be carried out.

[0120] On the other hand, a privacy data security protection system for the manufacturing industry based on blockchain is implemented based on the foregoing privacy data security protection system for the manufacturing industry based on blockchain, as Figure 1 shown, and includes the following steps:

[0121] Step 1: Collect the privacy data of the manufacturing industry;

[0122] Mobile devices are used to collect industrial data from intelligent instruments and are the basis for providing advanced intelligent industrial services. Mobile intelligent devices include different types of devices, such as monitors, sensors, and transmission devices. After obtaining permission, local edge service providers, on the one hand, allow mobile devices to offload data locally to the blockchain network; on the other hand, they will package the data collected from the manufacturing industry (such as environmental data and historical data of mechanical operations) and participate in global blockchain data sharing.

[0123] Step 2: Make an offloading decision for the processing task of privacy data;

[0124] Suppose there are N randomly distributed mobile users Each user n Generates a computing task Tn = {Dn, Xn, Cn}. The computing tasks of users can be offloaded locally or to multiple mobile edge computing servers. Among them, D n Represents the total data volume of the computing task T of user n n , x n Represents the proportion of the computing task of mobile user n offloaded to the mobile edge computing on the edge server. Cn represents the number of mobile CPU cycles spent by user n to process unit data. After the user generates the computing task, an offloading decision is made to reasonably schedule the architecture resources for data calculation.

[0125] Step 2.1: Latency of data offloading

[0126] The latency includes: the latency of the data offloading edge computing part And the latency T' of the blockchain verification part during the offloading process n ; The latency of the data offloading edge computing part includes: the execution time of offloading to the local The processing time of offloading to the mobile edge computing server Among them Includes three parts: the task encryption time at the local end The time to send the task to the mobile edge computing server And the task execution time on the mobile edge computing server

[0127] Among them, D n Is the size / bit of the input data, Are the CPU utilization rates of the mobile / edge devices cycles / bit respectively, Refer to the CPU cycle frequencies of the local / edge devices cycles / s respectively. x n Refers to the proportion of data offloading to the edge server, Refers to the CPU utilization rate of the local device for encryption processing:

[0128]

[0129]

[0130] where w n is the proportion of the bandwidth resource allocated to the nth task, B is the total radio bandwidth, p n is the transmission power per unit time of the mobile user n, h n is the channel gain between the mobile device and the edge computing server, and N0 is the Gaussian white noise. R n is the uplink transmission rate, defined in bit / s as:

[0131]

[0132] The latency T' of the blockchain verification part during the offloading process n respectively includes: 1) the time for the blockchain manager to send the unverified block 2) the time for block verification 3) the broadcast and comparison of the verification results ψ(n'*B)m' among the verifiers; 4) the time to send the verification feedback

[0133]

[0134] where m' is the number of selected verifiers, with the maximum and minimum values being v and M respectively, n' is the number of transactions per block, with the maximum and minimum values being x and t respectively, B is the transaction size, K is the computing resource required for the block verification task, x i is the available computing resource at the verifier i, O is the size of the verification feedback, R n and r u are the downlink and uplink transmission rates from the blockchain manager to the verifier respectively. ψ is a predefined parameter obtained using the statistics of block verification;

[0135] Since the task T of user n n can be offloaded to the local and mobile edge servers, the total latency for simultaneously processing the data at the local and mobile edge servers to complete the entire offloading task is:

[0136]

[0137] Step 2.2: Energy consumption of data offloading;

[0138] The energy consumption E n includes the energy consumption of the data offloading edge computing part and the energy consumption of the blockchain verification part during the offloading process The energy consumption of the data offloading edge computing part includes local energy consumption The energy consumption of offloading to mobile edge computing for processing Among which includes encryption energy consumption Transmission energy consumption And computing energy consumption

[0139]

[0140]

[0141]

[0142]

[0143]

[0144]

[0145] Among them, P M ,P sec ,P T ,P C are respectively the energy consumed per unit time for local data processing, local data encryption, data transmission, and edge data computing.

[0146] In a blockchain, such as Ethereum, all transactions have a cost measured in gas units, regarded as a measure of the standardized contract cost. The energy consumption of the blockchain verification part during the offloading process Among them, is the maximum amount of gas for executing a transaction, ξ is the natural gas price value, the ratio paid to miners per unit of natural gas; g n is the amount of gas used when executing the access control contract, and Ek refers to the energy consumed per unit of natural gas generated:

[0147]

[0148] Step 2.3 Normalization processing:

[0149]

[0150] Among which α and β are respectively the proportions of each part, and α + β = 1, are respectively the maximum values of each part, i.e., latency and energy.

[0151] Step 2.4 Formulate the objective function:

[0152] Formulate the multi-user secure offloading joint optimization offloading decision A, blockchain configuration B = {m′, n′}, computing resource allocation f = [f1, f2,... fN and the allocation of radio bandwidth of communication resources w = [w1, w2,..., w N and the use of smart contracts:

[0153]

[0154] Constraints:

[0155] (C1) x n ∈ [0, 1] n ∈ N′,

[0156] (C2) v ≤ m′ ≤ M,

[0157] (C3) t ≤ n′ ≤ x,

[0158] (C4)

[0159] (C5)

[0160] (C6) 0 < w n ≤ 1, n ∈ N′,

[0161] (C7)

[0162] (C8)

[0163] Among them, (C1) - specific offloading allocation constraints; C(2), C(3) - blockchain configuration constraints; C(4), C(5) - mobile edge computing computing power constraints, computing resource constraints; C(6), C(7) bandwidth allocation constraints; C(8) - smart contract constraints.

[0164] Step 2.5: For the data offloading problem, an advanced deep reinforcement learning offloading (ADRLO) scheme is designed, especially considering minimizing costs. Specifically, ADRLO is a multi-stage scheme, including two stages:

[0165] (1) The data security supervision process on the blockchain based on smart contract technology; A specific protocol for smart contract data offloading is designed, and data offloading is securely executed according to the protocol monitoring architecture.

[0166] (2) Data computing offloading algorithm based on deep reinforcement learning (DRL). An algorithm based on asynchronous advantage actor-critic is proposed, which updates the architecture parameters of medical data offloading by considering the resource allocation of each user and minimizing the total cost L. In the following subsections, the security audit design based on smart contracts is first introduced. Then, the data offloading scheme based on deep reinforcement learning is presented.

[0167] Step 2.5.1: Data security supervision process on the blockchain based on smart contract technology

[0168] For all data offloaded to the edge, we must ensure its data security during transmission and conduct an audit of the data, which we perform on the smart contract.

[0169] For the enterprise's data, the enterprise first uses the Hash function to generate a hash table of the data and encrypts the hash table with its own private key to generate a digital signature, and then uploads the data. The digital signature and the encrypted data are sent to the industrial data service provider together; after receiving the data, the industrial data service provider decrypts the digital signature with the enterprise's public key. According to this method, the hash table of the encrypted data is obtained; the industrial data service provider uses the Hash function on the data and compares the obtained data with the digital signature in the previous step; if the two are the same, it means that the data has not been modified and the privacy of the encrypted data has not been compromised.

[0170] Step 2.5.1.1: The enterprise registers, and the privacy data security protection system designs public and private keys for the enterprise; the public and private keys are {a, b, u, f, spk, v} and {ω, q, ssk} respectively, where ω, q are random numbers, spk and ssk are randomly generated key pairs, v ∈ L1, u, f ∈ L2, and L1 and L2 are multiplicative cyclic groups. v, u, f are the generators of the multiplicative cyclic group, and the random numbers a, b satisfy the following equations:

[0171] v = (ω + q) 2 (a + b) ω+q

[0172] Step 2.5.1.2: According to the offloading decision, the enterprise divides the data offloaded to the mobile edge computing server into J blocks, denoted as the set S, S = {m i}, i ∈ {1, 2,..., J}, where i refers to the subscript of the J data blocks. m i represents the i-th data block. The enterprise uses the private key {ω, q, ssk} to calculate the signatures of the J blocks of data, and represents R(m i ) as the index of the corresponding hash table after hashing, and the root of the hash tree is R MHT, obtain the processed data {S, Φ}, where the signature set Φ is defined as:

[0173]

[0174] where σ i represents the signature of the data block m i .

[0175] Step 2.5.1.3: Send the processed data {S, Φ} to the industrial data service provider for authorized data storage;

[0176] Processing by the industrial data service provider: After receiving the processed data, if the latest nonce value used to generate the random number is directly obtained from Ethereum; it is not secure because a malicious attacker controls the generation of the random number; in this case, for the security of industrial data, a delay function is adopted to prevent the service provider from controlling the generation of the random number when the industrial data service provider is untrusted; specifically, execute the delay function f time , obtain the seed seed for generating the random number, specifically as follows:

[0177] seed = f time (nonce new )

[0178] where time is the delay time and nonce new is a new nonce value; after executing the delay function, the smart contract will verify the time value at point J, and the specific data is the set {(i, f time (nonce new ))}, i ∈ {1, 2,..., J}; when the verification is successful, a challenge, that is, verification information, will be created; if the verification fails, this service will be rejected and a suitable seed set will be selected again.

[0179] The smart contract is as follows: when the system receives the data provided by the industrial data service provider, it will check its identity to determine whether its address is the same as the previously registered address; if the audit is successful, the corresponding challenge chal = {(i, v i )}, i ∈ {1, 2,..., J}, and send it to the industrial data service provider, where v i represents the values of J randomly selected seeds seed. If the audit is unsuccessful, the industrial data service provider will be punished.

[0180] Step 2.5.1.4: After receiving the challenge sent by the smart contract, the industrial data service provider calculates the proof;

[0181] Step 2.5.1.4.1: After receiving the challenge chal, the edge computing service provider selects random numbers l, h ∈ Z p , where Z p is the set of non - negative integers less than the constant p.

[0182] Step 2.5.1.4.2: The industrial data service provider calculates the proof Proof = {Rand, μ, σ, }, where Rand, μ, σ are the proof sub - information constructed for this proof

[0183] Rand = [(u + f) ω+q l , Rand ∈ L1

[0184]

[0185]

[0186]

[0187] Step 2.5.1.4.3: Send the proof Proof to the smart contract for verification.

[0188] Step 2.5.1.5: The smart contract verifies the proof generated by the industrial data service provider;

[0189] Create an object EnterprisePe = EnterpriseAddr(Addr) according to the enterprise address, initialize the variable sum = 0, traverse index from 0 to Pe.ChalNum.i.length, and calculate sum: sum = sum + Pe.ChalNum.i[index] * Pe.chal.v i [index]. Calculate and determine whether e(σ * Rand h*Rand , a + b) is equal to , where e(*, *) refers to the bilinear mapping; if they are equal, the verification result is successful and the data is allowed to be processed at the edge end, otherwise, the edge service provider is punished.

[0190] Step 2.5.2: Data calculation offloading algorithm based on deep reinforcement learning DRL

[0191] Propose the state space, action space, and reward function in the deep reinforcement learning offloading ADRLO, which are defined as a Markov decision process as follows:

[0192] 2) State space: The state space describes the remaining resource quantity:

[0193] ​· ec: Computing resources available to the mobile edge computing server

[0194] · bw: Bandwidth resources available for the industrial mobile edge computing server

[0195] · ag: Maximum gas limit for security supervision provided by the blockchain audit contract

[0196] Therefore, the state of the architecture is represented as:

[0197] State = [ec, bw, ag]

[0198] 2) Action space: Define the action space as four vector sets, namely: offloading decision vector X, mobile edge computing server resource allocation vector G, radio resource bandwidth allocation vector w, and smart contract gas allocation vector set g:

[0199] · X = [X1,..., X n ,..., X N , where X n is the proportion of computing tasks assigned by enterprise n to the mobile edge computing server.

[0200] · G = [G1,..., G n ,..., G N , where G n is the size of the computing resources of mobile edge computing s for the tasks of enterprise n.

[0201] · w = [w1,..., w n ,..., w N , where w n represents the ratio of the wireless transmission rate allocated to enterprise n.

[0202] · g = [g1,..., g n ,..., g N , where g n is the amount of gas consumed for the security supervision of the audit contract of enterprise n.

[0203] Therefore, the action of the system is represented as:

[0204] Action = [X, G, w, g] = [X1, G1, w1, g1,..., X N , G N , w N , g N

[0205] ​3) Reward: The task of the agent is to continuously explore new actions in the environment to obtain the maximum reward. The agent searches for the optimal offloading operation for the industrial data in each state. The goal is to minimize the total cost in the architecture, namely the task computing cost and the blockchain verification cost. Therefore, our reward is related to the formulated objective function. In our system, to minimize the total cost \(C(s, a)\) of the latency cost \(T(s, a)\) and the energy cost \(E(s, a)\), the reward is set as:

[0206] \(r(s, a)= -C(s, a)=-(T(s, a)+E(s, a))=-L\)

[0207] 4) ADRLO Algorithm: The ADRLO algorithm is an Actor-Critic learning method that combines policy gradient and temporal difference learning. Actor refers to the policy function \(\pi\) θ (a|s), that is, the actor will learn the policy to obtain the highest reward. Critic refers to the value function \(V\) π (s), and the value function is used to achieve the single-step update of the algorithm. In the policy gradient, we update the parameter \(\theta\), and calculate the gradient as follows:

[0208]

[0209] where \(H\) represents the number of \(\tau\) and the random number \(\tau\in[0, 1]\), \(b'\) is the baseline, \(T'\) n represents the time period, is to discount the future reward, \(\gamma\) t′-t is the discount factor. refers to the future reward. refers to the update of the action control parameter \(\theta\) of the agent, that is, given an \(s\) t , what kind of \(a\) the actor wants to take t will depend on the actor's parameter \(\theta\), so this part can be controlled by the actor himself. Use the expected value of the cumulative reward to replace the reward from \(t\) to \(T'\) n time period: refers to the Q function, that is, the evaluation network, \(E[\cdot]\) is the expected value, and use the value function to replace the baseline \(b\), does not involve actions, involves actions, is 's expected value. Therefore, and are positive or negative numbers, and the Actor-critic calculates the gradient as follows.

[0210]

[0211] However, the above Actor - critic needs to estimate the Q - network and V - network, which may lead to inaccurate estimation. Therefore, we use the value of the V - network to replace the value of the Q - network, obtain the reward r, and jump to the state r by taking the action a in the state s t+1 . The expected value of is equal to the Q - function:

[0212]

[0213] To further approximate the expected value, replace the Q - function with That is, the current reward plus the value function of the next state In the ADRLO environment, a participant interacts with the environment through the policy π to generate data. After collecting the data in the policy gradient method, the policy is updated. First, use these data to estimate the value function, and use temporal difference or Monte Carlo to estimate the value function. Update the policy π according to the value function. The ADRLO calculates the gradient as follows:

[0214]

[0215] After obtaining π, interact with the environment to collect new data, estimate the value function, and use the new value function to update the policy and actor.

[0216] To improve the efficiency of the proposed ADRLO algorithm, we consider the asynchronous idea and assign tasks to multiple threads simultaneously. Each thread updates the learned parameters to the global network and pulls the global parameters for the next learning. In this case, we propose the ADRLO algorithm to train the parameters of the industrial data offloading problem.

[0217] In this embodiment, we first evaluate the proposed industrial data security audit. Specifically, the enterprise device is executed by a mobile phone configured with Redmi K30, Qualcomm Snapdragon 765G processor, and 6GB RAM. The industrial data service provider is carried out on a computer configured with AMD Ryzen 5 4600G, 16GB RAM, Radeon Graphics, and 3.70GHz processor. Then, we quickly generate virtual private chain nodes for these edge devices and local devices through ganache software for development and testing. Next, we configure the environment using Remix IDE and connect to the virtual private chain using Web3 Provider. We write smart contracts through Solidity and deploy them on the local simulation network and the official Ethereum test network Ropsten. Finally, when performing data offloading security verification, we operate these nodes according to the provisions of the smart contract.

[0218] Next, we present the specific work of each simulated role:

[0219] · Enterprise: First, we set relevant random numbers. Based on the ecdsa library tool in Python, we use the elliptic curve encryption algorithm to generate the public key and private key of the enterprise, completing the initialization of the enterprise. Then, we use the above-mentioned enterprise private key to simulate the signature of the data, which is later used to verify the legality of the transaction. In addition, we split the enterprise industrial data collected from the enterprise into blocks and use the SH3-keccak256 method to construct a Merkle hash tree (MHT) to encrypt the enterprise industrial data. After processing, we package and send the encrypted data together with the signature to the industrial data service provider. Through smart contract verification, the industrial data service provider is authorized to store it.

[0220] · Industrial Data Service Provider (IDSP): First, the IDSP plays a delay function. Specifically, we use the time library in Python, as well as the sleep and random functions in the random library. After randomly sleeping for a period of time, the IDSP saves the value at multiple time points. Then, the MSP sends these time points together with the corresponding values to the smart contract for verification. After the smart contract verification is successful, the verification information is sent to the IDSP. Finally, the IDSP calculates the corresponding proof Proof = {Rand, μ, σ, }, and sends the proof to the smart contract for verification. Successful authentication authorizes it to process and store enterprise data.

[0221] · Smart Contract: First, we securely verify the transaction between the enterprise and IDSPs. After the client sends the packaged data, the smart contract verifies the identity of the IDSP based on the signature to ensure the legality of the transaction. Secondly, after the smart contract receives the IDSP, it calculates the proof Check the proof to determine data security.

[0222] 1) Use Case Evaluation

[0223] This patent conducts test experiments on the security, latency, energy consumption, and total cost of the system.

[0224] 2) Baseline Algorithm

[0225] This patent compares the passing rates of the tests in terms of security with the multi-point detection audit scheme and the no-detection scheme. During the test, we simulated attacks. These attacks would cause packet loss or increase (i.e., change the number of packets), and would also cause the content of the packets to be tampered with. We judged whether the attacks were detected based on the number of audits passed in the final test. The lower the number of passes, the higher the security.

[0226] This patent also tested the performance of our model in terms of latency, energy consumption, and total cost. The baseline schemes for comparison include: Double-DQN, DDPG, DQN, EO, LO, Random. Among them, DQN also uses the DRL algorithm for resource deployment in the computing offloading task. DDPG integrates a deep neural network into the deterministic policy gradient (the state space and action space are continuous, and the exploration space is infinite) to solve the resource deployment problem, while Double DQN performs resource deployment through Double DQN with a continuous state space and a discrete action space. Edge offloading (EO) offloads all tasks to the MEC server for execution. For the resource allocation in the industrial data offloading process, the mean value is taken. Local offloading (LO) processes all task executions in the local device. In the Random method, the computing offloading and resource deployment randomly select resources from the MEC server and the local device. The smaller the latency, energy consumption, and total cost obtained from the experiment, the better.

[0227] 3) Evaluation results

[0228] In this patent, we evaluated the number of detection attacks passed, latency, energy consumption, and total cost of the experiment. The specific experimental results are as Figure 3 and Figure 4 shown. Among them, Figure 3 (a) is a latency comparison graph of local data encryption and batch auditing of smart contracts, Figure 3 (b) is a latency comparison graph of verifying signatures and proofs of smart contracts, Figure 3 (c) is a comparison graph of the number of detections passed by our method, multivariate point audit detection, and no-verification algorithms in the data processing containing attacks. Figure 4 (a) is a comparison graph of the latency consumption of seven algorithms (LO, Random, EO, DDPG, DQN, Double DQN, ADRLO) during data offloading at different time periods, Figure 4 (b) is a comparison graph of the energy consumption of seven algorithms (LO, Random, EO, DDPG, DQN, Double DQN, ADRLO) during data offloading at different time periods, Figure 4 (c) is a comparison graph of the total cost consumption of seven algorithms (LO, Random, EO, DDPG, DQN, Double DQN, ADRLO) during data offloading at different time periods.

[0229] The present invention is superior to the baseline model in the above four aspects, demonstrating the good effects of the present invention in terms of security performance, latency, and energy consumption. The present invention demonstrates the security of the mechanism based on smart contracts. In addition, it has significant advantages over the baseline algorithm in reducing latency and energy consumption, indicating the feasibility of the present invention's solution in industrial data offloading applications.

[0230] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles applied. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by the specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the technical solutions formed by mutually replacing the above features with the technical features (but not limited to) disclosed in the embodiments of the present disclosure that have similar functions.

Claims

1. A method for protecting the security of privacy data in the manufacturing industry based on blockchain, characterized in that, It includes the following steps: Step 1: Collect privacy data of the manufacturing industry; The mobile device is used to collect industrial data from intelligent devices. After obtaining permission, on the one hand, the local edge service provider allows the mobile device to offload data locally to the blockchain network; on the other hand, it will package the data collected from the manufacturing industry and participate in the global blockchain data sharing; Step 2: Make an offloading decision for the processing task of privacy data, and audit the data through blockchain smart contract technology to ensure the security of the data offloaded to the edge side; specifically, for the data offloading problem, a deep reinforcement learning offloading ADRLO scheme is designed; The deep reinforcement learning offloading ADRLO scheme specifically includes: the data security supervision process on the blockchain based on smart contract technology, and the data computing offloading algorithm based on deep reinforcement learning DRL; Among them, the data computing offloading algorithm based on deep reinforcement learning DRL defines the state space, action space and reward function in deep reinforcement learning offloading ADRLO as a Markov decision process as follows: (1) State space: The state space describes the amount of remaining resources: ec: Computing resources available to the mobile edge computing server bw: Bandwidth resources for industrial mobile edge computing servers ag: Maximum gas limit for security supervision provided by the blockchain audit contract Therefore, the state of the architecture is represented as: State = [ec, bw, ag] (2) Action space: The action space is defined as four vector sets, namely: offloading decision vector X, mobile edge computing server resource allocation vector G, radio resource bandwidth allocation vector w, smart contract gas allocation vector set g: X = [X1,..., X n ,..., X N , where X n is the proportion of computing tasks allocated by enterprise n to the mobile edge computing server; G = [G1,..., G n ,..., G N , where G n is the size of the mobile edge computing s computing resources for the tasks of enterprise n; w = [w1,..., w n ,..., w N , where w n represents the ratio of the wireless transmission rate allocated to enterprise n; g = [g1,..., g n ,..., g N , where g n is the gas consumption for the audit contract security supervision of enterprise n; Therefore, the action of the system is represented as: Action = [X, G, w, g] = [X1, G1, w1, g1,..., X N , G N , w N , g N ​ (3) Reward: The task of the agent is to continuously explore new actions in the environment to obtain the maximum reward. The agent finds the optimal offloading operation for the industrial data in each state; the goal is to minimize the total cost in the architecture, that is, the task computing cost and the blockchain verification cost; minimize the total cost C(s,a) of the latency cost T(s,a) and the energy cost E(s,a), and set the reward as: r(s,a) = -C(s,a) = -(T(s,a) + E(s,a)) = -L 4) ADRLO algorithm: Policy function π θ (a|s), that is, the actor will learn the policy to obtain the highest reward.

2. The method for protecting the privacy data security of the manufacturing industry based on blockchain according to claim 1, wherein, In step 2, assume there are N randomly distributed mobile users Each user generates a computing task Tn = {Dn, Xn, Cn}, where D n represents the total data volume of the computing task T n of user n, x n represents the proportion of the computing task of mobile user n offloaded to the mobile edge computing of the edge server, and Cn represents the number of mobile CPU cycles spent by user n to process unit data.

3. A method for protecting the security of privacy data in the manufacturing industry based on blockchain according to claim 1, characterized in that, Step 2 specifically includes the following steps: Step 2.1: Calculate the latency of data offloading; Step 2.2: Calculate the energy consumption of data offloading; Step 2.3 Normalization processing: where α and β are the proportions of each part respectively, and α + β = 1, are each part, namely the maximum values of delay and energy respectively; Step 2.4 Formulate the objective function: Formulate a joint optimization offloading decision for multi-user secure offloading A, blockchain configuration B = {m ′ , n ′}, computing resource allocation f = [f1, f2,..., f N and allocation of communication resource radio bandwidth w = [w1, w2,..., w N , as well as the use of smart contracts. The objective function is as follows: Constraints: (C1)x n ∈[0,1] n ∈ N′, (C2) v ≤ m ′ ≤ M, (C3)t ≤ n ′ ≤ x, (C6)0<w n ≤1, n ∈ N′, Among them, (C1) - specific offloading allocation constraints; C(2), C(3) - blockchain configuration constraints; C(4), C(5) - mobile edge computing computing power constraints, computing resource constraints; C(6), C(7) bandwidth allocation constraints; C(8) - smart contract constraints; Step 2.5: For the data offloading problem, a deep reinforcement learning offloading ADRLO scheme is designed.

4. The method for protecting the privacy data security of the manufacturing industry based on blockchain according to claim 3, characterized in that The time delay described in Step 2.1 includes: the time delay of the data offloading edge computing part and the time delay T of the blockchain verification part during the offloading process n ′ ; the time delay of the data offloading edge computing part includes: the execution time of offloading to the local the time of offloading to the mobile edge computing server for processing where it includes three parts: the task encryption time at the local end the time of sending the task to the mobile edge computing server and the task execution time on the mobile edge computing server Among them, D n is the size / bit of the input data, is the CPU utilization cycles / bit of the mobile / edge device respectively, respectively refers to the CPU cycle frequency cycles / s of the local / edge device, x n refers to the proportion of data offloaded to the edge server, refers to the CPU utilization of the local device for encryption processing: where, w n is the proportion of the bandwidth resource allocated to the nth task, B is the total radio bandwidth, p n is the transmission power per unit time of mobile user n, h n is the channel gain between the mobile device and the edge computing server, N0 is the Gaussian white noise; R n is the uplink transmission rate, in units of bit / s and is defined as: The latency T of the blockchain verification part during the uninstallation process n ′ respectively include: 1) the time for the blockchain manager to send the unverified blocks 2) the time for block verification 3) the verification result broadcast and comparison ψ(n ′ *B)m ′ ; 4) the time for sending verification feedback where m ′ is the number of selected validators, with the maximum and minimum values being v and M respectively, and n ′ is the number of transactions per block, with the maximum and minimum values being x and t respectively, B is the transaction size, K is the computing resource required for block verification tasks, and x i is the computing resource available at validator i, O is the size of the verification feedback, and R n and r u are the downstream and upstream transmission rates from the blockchain manager to the validator respectively; ψ is a predefined parameter obtained using the statistics of block verification; Due to the task T of user n n which can be offloaded to local and mobile edge servers, the total latency to complete the entire offloading task while processing local and mobile edge server data is:

5. A method for protecting the security of privacy data in the manufacturing industry based on blockchain according to claim 3, wherein, The energy consumption E described in Step 2.2 n includes the energy consumption of the data offloading edge computing part and the energy consumption of the blockchain verification part during the offloading process The energy consumption of the data offloading edge computing part includes local energy consumption the energy consumption of offloading to mobile edge computing for processing which includes encryption energy consumption transmission energy consumption and computing energy consumption Among them, P M , P sec , P T , P C are the energy consumed per unit time for local data processing, local data encryption, data transmission, and edge data computing respectively; In the blockchain, all transactions have a cost measured in gas units, which is considered a measure of the cost of a standardized contract, and the energy consumption of the blockchain verification part during the offloading process Among them, is the maximum amount of gas for executing a transaction, ξ is the natural gas price value, the ratio paid to miners per unit of natural gas; g n is the amount of gas used when executing the access control contract, and Ek refers to the energy consumed to produce a unit of natural gas:

6. The method for protecting the privacy data security of the manufacturing industry based on blockchain according to claim 3, wherein Step 2.5 specifically includes the following steps: Step 2.5.1: The data security supervision process on the blockchain based on smart contract technology For the enterprise's data, the enterprise first uses a Hash function to generate a hash table of the data, and uses its own private key to encrypt the hash table to generate a digital signature. Then, the data is uploaded, and the digital signature and the encrypted data are sent to the industrial data service provider together. After receiving the data, the industrial data service provider decrypts the digital signature using the enterprise's public key, obtains the hash table of the encrypted data according to this method, and the industrial data service provider uses the Hash function on the data and compares the obtained data with the digital signature in the previous step. If the two are consistent, it means that the data has not been modified and the privacy of the encrypted data has not been compromised. Step 2.5.1.1: The enterprise registers, and the privacy data security protection system designs public and private keys for the enterprise. The public and private keys are {a, b, u, f, spk, v} and {ω, q, ssk} respectively, where ω and q are random numbers, spk and ssk are randomly generated key pairs, v ∈ L1, u, f ∈ L2, and L1 and L2 are multiplicative cyclic groups. v, u, and f are the generators of the multiplicative cyclic group, and the random numbers a and b satisfy the following equations: v = (ω + q) 2 (a + b) ω+q Step 2.5.1.2: According to the offloading decision, the enterprise divides the data offloaded to the mobile edge computing server into J blocks, denoted as set S, S = {m i}, i ∈ {1, 2,..., J}, where i refers to the subscript of the J data blocks; m i represents the i-th data block; the enterprise uses the private key {ω, q, ssk} to calculate the signatures of the J blocks of data, and represents R(m i ) as the index of the corresponding hash table after hashing, then the root of the hash tree is R MHT , and the processed data {S, Φ} is obtained, where the signature set Φ is defined as: where, σ i represents the signature of data block m i ; Step 2.5.1.3: Send the processed data {S, Φ} to the industrial data service provider for authorized data storage. Industrial data service provider processing: After the industrial data service provider receives the processed data, if the latest nonce value used to generate the random number is directly obtained from Ethereum; it is not secure because a malicious attacker controls the generation of the random number. In this case, a delay function is used to prevent the service provider from controlling the generation of the random number when the industrial data service provider is untrusted; specifically, the delay function f is executed time , to obtain the seed seed for generating the random number, as follows: seed=f time (nonce new ) where time is the delay time and nonce new is a new nonce value; after executing the delay function, the smart contract will verify the time value at point J, and the specific data is the set {(i, f time (nonce new ))}, where i ∈ {1, 2, …, J}; when the verification is successful, a challenge, that is, the verification information, will be created; if the verification fails, this service will be rejected and a suitable seed set will be selected again; The smart contract is as follows: when the system receives data provided by an industrial data service provider, it will check its identity to determine whether its address is the same as the previously registered address; if the audit is successful, the corresponding challenge chal = {(i, v i )}, where i ∈ {1, 2, …, J}, and it will be sent to the industrial data service provider, where v i represents the values of J randomly selected seeds seed; if the audit is unsuccessful, the industrial data service provider will be punished; Step 2.5.1.4: After receiving the challenge sent by the smart contract, the industrial data service provider calculates the proof. Step 2.5.1.4.1: After receiving the challenge chal, the edge computing service provider selects random numbers l, h ∈ Z p , where Z p is the set of non-negative integers less than the constant p; Step 2.5.1.4.2: The industrial data service provider calculates the proof according to the following formula where Rand, μ, and σ are the proof sub-information constructed for this proof respectively Rand = [(u + f) ω+q l , Rand ∈ L1​ Step 2.5.1.4.3: Send the proof Proof to the smart contract for verification. Step 2.5.1.5: The smart contract verifies the proof generated by the industrial data service provider. Create an object EnterprisePe = EnterpriseAddr(Addr) according to the enterprise address, initialize the variable sum = 0, traverse index from 0 to Pe.ChalNum.i.length, and calculate sum: sum = sum + Pe.ChalNum.i[index] * Pe.chal.v i [index]; calculate and judge e(σ * Rand h*Rand , a + b) and whether they are equal, where e(*, *) refers to the bilinear mapping; if they are equal, the verification result is successful and the data is allowed to be processed at the edge, otherwise, punish the edge service provider; Step 2.5.2: The data calculation offloading algorithm based on deep reinforcement learning DRL. The ADRLO algorithm further includes: a value function V π (s), using the value function to achieve single-step update of the algorithm. In the policy gradient, update the parameter θ, and calculate the gradient as follows: where H represents the number of τ and the random number τ ∈ [0, 1], b ′ is the baseline, T n ′ represents the time period, is to discount the future rewards, γ t′-t is the discount factor; refers to the future rewards; refers to the update of the action control parameter θ of the agent, using the expected value of the cumulative rewards to replace the rewards from time t to T′ n time period: refers to the Q function, i.e., the evaluation network, E[*] is the expected value, using the value function to replace the baseline b, does not involve actions, involves actions, is the expected value of; Therefore, and are positive or negative, and the Actor - critic calculates the gradient as follows; Replace the value of the Q-network with the value of the V-network, obtain the reward r, and jump to the state r by taking the action a in the state s t+1 ; The expected value of is equal to the Q function: Getting closer to the expected value, replace the Q function with r + V: That is, the current reward plus the value function of the next state In the ADRLO environment, a participant interacts with the environment through the policy π to generate data. After collecting the data in the policy gradient method, the policy is updated. First, these data are used to estimate the value function, and the time series difference or Monte Carlo method is used to estimate the value function. The policy π is updated according to the value function. The ADRLO calculates the gradient as follows: After obtaining π, interact with the environment to collect new data, estimate the value function, and use the new value function to update the policy and actor.

7. The method for protecting the security of private data in the manufacturing industry based on blockchain is implemented based on a system for protecting the security of private data in the manufacturing industry based on blockchain, and is characterized in that, It includes a data collection module, a data offloading decision module, a data local processing module, a blockchain smart contract audit module, and a data edge processing module. The data collection module is used to collect the data of each enterprise, that is, the task processing information. The data offloading decision module receives the task processing information, formulates the cost problem as a Markov decision process, and solves the data offloading decision problem through a deep reinforcement learning method based on policy search, specifically including load distribution decision, allocation of computing resources and radio transmission bandwidth, and blockchain data security audit. For the information that the data local processing module finally decides to process locally by the data offloading decision module, it calls the local server for processing. For the information that the blockchain smart contract audit module finally decides to process at the edge by the offloading decision module, it counts the data before the data is transmitted to the edge, and conducts a smart contract audit on the data when it reaches the edge to check that there is no modification or replacement of the data and no loss or addition. For the information that the data edge processing module finally decides to process at the edge by the data offloading decision module, it calls the server at the edge for processing.

Images