A secure processing method for industrial Internet data

Abstract

The present invention relates to a method for securely processing industrial Internet data, and belongs to the field of industrial Internet data processing. The method comprises: a factory and a collaborating party generate a pair of secret keys using an ElGamal encryption algorithm; the contracting party initializes model parameters and uploads them to a blockchain with a public key; the factory downloads the initial model and the collaborating party's public key from the blockchain, trains the model, and extracts model parameters using a differential privacy algorithm; the factory encrypts the model parameters and stores them in IPFS to obtain a hash value; the encrypted hash value is added to the blockchain with the factory public key; the collaborating party retrieves the model parameters in IPFS through the hash value to train the global model, uses SK to encrypt the model parameters and stores them in IPFS, encrypts the IPFS hash value, encrypts SK using the factory public key, and adds the result to the blockchain; the factory receives the current global model parameters for update. The present invention solves the privacy and trust issues of machine learning in industrial Internet systems.

Description

Technical Field

[0001] The present invention belongs to the technical field of industrial Internet data security processing and relates to a method for securely processing industrial Internet data. Background Art

[0002] In recent years, with the large-scale deployment and application of intelligent sensing devices, massive amounts of data have been generated. These data contain information from all areas of production and life and have become an important production factor. As a product of the deep integration of new-generation information technology and manufacturing, the Industrial Internet builds a new industrial production and service system that is fully connected with all factors, the entire industrial chain, and the entire value chain through the comprehensive interconnection of people, machines, and objects. It is a way to achieve digital transformation and a key force in achieving the transformation of new and old kinetic energy. The Industrial Internet is in a new stage of explosive growth in data scale, dramatic increase in computing power, and continuous improvement in algorithm performance. Intelligent technologies such as deep learning, reinforcement learning, and federated learning have become an important support for improving the comprehensive service performance of the Industrial Internet.

[0003] Industrial production data is highly confidential. The currently commonly used centralized model training methods have data privacy leakage problems, and it is difficult to take into account the dual needs of data sharing and privacy protection. Not only a new privacy-preserving computing paradigm is needed to complete multi-party joint modeling, but also a flexible and intelligent adaptation method is needed to ensure the efficient operation of the new computing paradigm. And security is the guarantee of the Industrial Internet. In the future, the construction of the emerging infrastructure of the Industrial Internet will continue to advance to a wider range, deeper degree, and higher level. Faced with new technologies and new security protection situations, the security ecology of the Industrial Internet will continue to face new challenges. Therefore, there is an urgent need for secure processing methods for Industrial Internet data to enhance the ability to deal with security risks and promote the prosperity and development of the Industrial Internet. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a secure processing method for industrial Internet data. In order to address the dual needs of industrial data sharing and privacy protection that are difficult to balance, by combining federated learning, differential privacy, ElGamal encryption, etc., a secure industrial Internet is collaboratively created to better empower the industrial Internet and promote industrial upgrading.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] A method for securely processing industrial Internet data, specifically comprising the following steps:

[0007] S1: The factory and the collaborator generate a pair of secret keys; then the collaborator initializes the model parameter w0 and uploads the model parameter w0 and its public key to the blockchain;

[0008] S2: Each factory downloads the initial model parameter w0 and the public key of the collaborator from the blockchain, and uses a specific number of local data samples to train a deep neural network model. After the training is completed, the model is processed using a differential privacy algorithm to generate a local differential privacy machine learning model, and then the factory extracts the model parameters.

[0009] S3: The factory uses the public key of the collaborator to encrypt the model parameters, then stores the encrypted model parameters in IPFS and obtains a unique IPFS hash value; then the IPFS hash value is also encrypted using the public key of the collaborator, and is packaged with the factory public key and added to the blockchain through a smart contract; the factory notifies the collaborator that the current round of local model training is complete;

[0010] The encrypted model parameters are stored in IPFS, which is a peer-to-peer network protocol for file storage that uses content-based addressing rather than location-based addressing. Files in the IPFS network are assigned a hash value, which is similar to a "fingerprint" and is calculated from the file content.

[0011] S4: The collaborating party uses the private key to decrypt, encrypts the IPFS hash value with its public key, retrieves the corresponding encrypted model parameters in IPFS using the IPFS hash value, and trains the global model after decryption and extracts the model parameters;

[0012] S5: Use SK (a symmetric key generated by the collaborating party) to encrypt the model parameters and store them in IPFS. After that, an IPFS hash value is assigned. The IPFS hash value is encrypted using SK. At the same time, the SK is encrypted using the public key of each factory. The encrypted result is added to the blockchain (this multi-layer encryption method ensures data security) and the factory is notified that this round of aggregation is over.

[0013] S6: After receiving notification from the collaborating party that the current federation cycle has ended, the factory retrieves the encrypted global model parameters, decrypts them using its private key to obtain SK, decrypts the encrypted IPFS hash value using SK, and retrieves the model parameters using the IPFS hash value; the factory uses the updated model parameters for local training, starts the next round of joint training, and repeats all steps in the predefined number of joint rounds.

[0014] Further, in step S1, the factory and the collaborating party generate a pair of secret keys using the ElGamal encryption algorithm; the ElGamal encryption algorithm is an asymmetric encryption algorithm based on the Diffie-Hellman key exchange. The ElGamal encryption algorithm is based on the principle that it is difficult to solve discrete logarithms, but its inverse operation can be effectively calculated by applying the square multiplication method. In the corresponding group G, the exponential function is a one-way function.

[0015] Further, step S2 specifically includes: the factory uses a specific number of local data samples to train a deep neural network; obtains the latest model parameters from the server, from 1 to the batch number Batch number b, calculate batch gradient g k (b) , locally update the model parameters: Where η represents the learning rate, w t represents the current model parameters, D k represents the dataset owned by the k-th collaborator, M represents the size of the mini-batch used for the specified customer update;

[0016] During the training process, differential privacy-based neural network training is implemented in SGD calculations by minimizing the empirical loss function To train the model parameters; at each step of SGD, the gradient is calculated, for the sampled subset, the norm l2 of each gradient is clipped, the average is calculated, noise is added to protect privacy, and a step is taken in the opposite direction of the average noise gradient to perform gradient descent back propagation to complete the training and finally output the model.

[0017] Further, step S4 specifically includes: Assuming that there are K participants (i.e., factories) in a federated learning system, D k represents the data set owned by the kth participant, P k represents the index set of data points located at participant k; let n k Indicates P k The cardinality of , assuming that the kth participant has n k When there are K participants in total, the collaborators aggregate the received model parameters, that is, perform weighted average on the received model parameters:

[0018]

[0019] Update model parameters:

[0020]

[0021] Among them, It means that the sample (x i ,y i ) is the loss result obtained by prediction, x i and i They represent the i-th training data point and its associated label respectively. η represents the learning rate, and n represents the number of training data. The collaborator checks whether the loss function converges or whether the maximum number of training rounds is reached. If so, the collaborator sends a signal to all participants to stop model training.

[0022] Further, in step S6, the model parameters updated by the factory are encrypted, stored and uploaded.

[0023] The beneficial effects of the present invention are: the present invention enhances the privacy and credibility of industrial Internet data by combining differential privacy, federated learning, blockchain and smart contracts, improves the ability to deal with security risks, and promotes the prosperity and development of the industrial Internet.

[0024] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0026] Figure 1 This is an architecture diagram of the industrial Internet data security processing method involved in the present invention;

[0027] Figure 2 A flow chart of the factory data processing process of the industrial Internet data security processing method of the present invention;

[0028] Figure 3 This is a flow chart of the collaborative data processing process of the industrial Internet data security processing method of the present invention. DETAILED DESCRIPTION

[0029] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0030] See also Figure 1 to Figure 3 , Figure 1 The following is an architecture diagram of an industrial Internet data security processing method. Figure 2 The figure shows the factory data processing process of the industrial Internet data security processing method. Figure 3 The figure shows the data processing process of the collaborative parties in the industrial Internet data security processing method.

[0031] This method solves the privacy and trust issues of machine learning in industrial Internet systems, including factories, collaborators, IPFS, blockchain and other entities and data transmission between entities. It specifically includes the following steps:

[0032] Step 1: The factory and the collaborator first generate a pair of secret keys using the ElGamal encryption algorithm. The ElGamal encryption algorithm is an asymmetric encryption algorithm based on the Diffie-Hellman key exchange. The ElGamal encryption algorithm is based on the principle that it is difficult to solve discrete logarithms, but its inverse operation can be effectively calculated using the square multiplication method. In the corresponding group G, the exponential function is a one-way function.

[0033] Step 2: Each factory downloads the initialization model w0 and the public key of the collaborator from the blockchain and trains a deep neural network using a specific number of local data samples. After training, the model is processed using a differential privacy algorithm to generate a local differential privacy machine learning model.

[0034] The factory trains a deep neural network using a specific number of local data samples. Get the latest model parameters from the server, from 1 to the batch size Batch number b, calculate batch gradient g k (b) , locally update the model parameters: η represents the learning rate.

[0035] During the training process, differential privacy-based neural network training is implemented in SGD calculations by minimizing the empirical loss function To train the model parameters. At each step of SGD, the gradient is calculated, and for the sampled subset, the norm l2 of each gradient is clipped, the average is calculated, noise is added to protect privacy, and a step is taken in the opposite direction of the average noise gradient to perform gradient descent back propagation to complete the training and finally output the model.

[0036] Federated learning can train deep learning models with the help of a central server while keeping the training data distributed on the client. The client only needs to submit its local gradient to the cloud server during model training, which to a certain extent avoids the risk of user data privacy leakage. Differential privacy is an efficient privacy protection technology that can quantify the degree of data privacy protection, set a suitable privacy budget, and achieve a good balance between data availability and privacy protection. Differential privacy is often used in the privacy protection of training data for artificial intelligence models, and can provide privacy protection for federated learning.

[0037] Under the premise that user data does not leave the local area, a shared model is established through parameter exchange and optimization under encryption mechanism or perturbation mechanism. The performance of this shared model is close to the model trained by aggregating data from all parties. This data joint modeling solution does not leak user privacy and complies with the principle of data security protection.

[0038] Step 3: The factory uses the public key of the collaborating party to encrypt the model parameters, then stores the encrypted model parameters in IPFS and obtains a unique IPFS hash value. The IPFS hash value is then encrypted using the public key of the contracting party, and is packaged with the factory public key and added to the blockchain using a smart contract. The factory notifies the collaborating party that the current round of local model training is complete.

[0039] The encrypted model parameters are stored in IPFS, which is a peer-to-peer network protocol for file storage that uses content-based addressing rather than location-based addressing. Files in the IPFS network are assigned a hash value, which is similar to a "fingerprint" and is calculated from the file content.

[0040] Step 4: The collaborating party uses the private key to decrypt the IPFS hash value encrypted by its public key, uses the IPFS hash value to retrieve the corresponding encrypted model parameters in IPFS, and after decryption, trains the global model and extracts the model parameters.

[0041] The collaborating party first decrypts to obtain the IPFS hash value, uses the IPFS hash value to retrieve the corresponding encrypted model parameters, and continues to use its private key to decrypt to obtain the parameters for the local model update.

[0042] Assume that there are K participants in a federated learning system, D k represents the data set owned by the Kth participant, P k represents the index set of data points located at customer k. Let n k Indicates P k The cardinality of , assuming that the kth participant has n k Data points, when there are K participants in total

[0043]

[0044] The collaborating party aggregates the received model parameters, that is, performs weighted average on the received model parameters:

[0045]

[0046] Where η represents the learning rate, and the collaborator checks whether the loss function converges or whether the maximum number of training rounds has been reached. If so, the collaborator sends a signal to all participants to stop model training.

[0047] Step 5: Use SK (a symmetric key generated by the collaborating party) to encrypt the model parameters and store them in IPFS. After that, an IPFS hash value will be assigned. Use SK to encrypt the IPFS hash value. At the same time, use the factory's public key to encrypt SK. The encrypted result is added to the blockchain to notify the factory that this round of aggregation is over.

[0048] Step 6: The factory uses the updated model parameters for local training. First, the encrypted global model parameters are retrieved, SK is obtained by decryption using the private key, the encrypted IPFS hash value is decrypted using SK, and the model parameters are retrieved using the IPFS hash value. After obtaining the model parameters, the model training is performed in the same way as in step 2. The model parameters are then encrypted, stored, and uploaded.

[0049] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A method for securely processing industrial Internet data, characterized in that: The method specifically comprises the following steps: S1: The factory and the collaborator generate a pair of secret keys; then the collaborator initializes the model parameter w0 and uploads the model parameter w0 and its public key to the blockchain; S2: Each factory downloads the initial model parameter w0 and the public key of the collaborator from the blockchain, and uses n local data samples to train a deep neural network model; after the training is completed, the model is processed with a differential privacy algorithm to generate a local differential privacy machine learning model, and then the factory extracts the model parameters, including: the factory uses n local data samples to train a deep neural network; obtains the latest model parameters from the server, from 1 to the batch size Batch number b, calculate batch gradient Update model parameters locally: Where η represents the learning rate, w t represents the current model parameters, D k represents the dataset owned by the k-th collaborator, M represents the size of the mini-batch used for the specified customer update; During the training process, differential privacy-based neural network training is implemented in SGD calculations by minimizing the empirical loss function To train the model parameters; at each step of SGD, the gradient is calculated, and for the sampling subset, the norm l2 of each gradient is clipped, the average is calculated, and a step is taken in the opposite direction of the average noise gradient to perform gradient descent back propagation to complete the training and finally output the model; S3: The factory uses the public key of the collaborator to encrypt the model parameters, then stores the encrypted model parameters in IPFS and obtains a unique IPFS hash value; then the IPFS hash value is also encrypted using the public key of the collaborator, and is packaged with the factory public key and added to the blockchain through a smart contract; the factory notifies the collaborator that the current round of local model training is complete; S4: The collaborating party uses the private key to decrypt, encrypts the IPFS hash value with its public key, retrieves the corresponding encrypted model parameters in IPFS using the IPFS hash value, and trains the global model after decryption and extracts the model parameters; S5: Use the symmetric key SK generated by the collaborating party to encrypt the model parameters and store them in IPFS. After that, an IPFS hash value will be assigned. The IPFS hash value is encrypted using SK. At the same time, the public key of each factory is used to encrypt SK respectively. The encrypted result is added to the blockchain to notify the factory that this round of aggregation has ended. S6: After receiving notification from the collaborating party that the current federation cycle has ended, the factory retrieves the encrypted global model parameters, decrypts them using its private key to obtain SK, decrypts the encrypted IPFS hash value using SK, and retrieves the model parameters using the IPFS hash value; the factory uses the updated model parameters for local training, starts the next round of joint training, and repeats all steps in the predefined number of joint rounds.

2. The method for securely processing industrial Internet data according to claim 1 is characterized in that: In step S1, the factory and the collaborating party generate a pair of secret keys using the ElGamal encryption algorithm; the ElGamal encryption algorithm is an asymmetric encryption algorithm based on the Diffie-Hellman key exchange.

3. The method for securely processing industrial Internet data according to claim 1, characterized in that: Step S4 specifically includes: Assuming there are K participants in a federated learning system, D k represents the data set owned by the kth participant, P k represents the index set of data points located at participant k; let n k Indicates P k The cardinality of , assuming that the kth participant has n k When there are K participants in total, the collaborators aggregate the received model parameters, that is, perform weighted average on the received model parameters: Update model parameters: Among them, It means that the sample (x i ,y i ) is the loss result obtained by prediction, x i and i They represent the i-th training data point and its related label respectively; η represents the learning rate, and n represents the number of training data; the collaborator checks whether the loss function converges or whether the maximum number of training rounds is reached; if so, the collaborator sends a signal to all participants to stop model training.

4. The method for securely processing industrial Internet data according to claim 1, characterized in that: In step S6, the model parameters updated by the factory are encrypted, stored and uploaded.

Images