A data transmission method, program, medium and device based on group signature and federated learning method
By adopting group signature-based data transfer method and differential privacy technology in federated learning, the problem of data source address exposure is solved, effective hiding and protection of data source address is realized, and the security of data transmission and the reliability of model are improved. .
Patent Information
- Application Number
- CN202410850920.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-28
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-06-28
AI Technical Summary
Existing federated learning methods have challenges in hiding data source addresses, resulting in the potential exposure of data source addresses, affecting data privacy protection.
Using a data delivery method based on group signature, the model update parameters are signed through group certificates and node private keys, and verified on the aggregation server side, generating group private keys and group certificates to protect the data source address, and further enhancing it with differential privacy methods Data privacy.
Effectively hide and protect the data source address, reduce the risk of data leakage, enhance the security and integrity of data transmission, ensure that the data participating in the learning has not been tampered with, and improve the reliability of model results.
Smart Images

Figure CN118396138B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of computer technology, and in particular to a data transmission method, program, medium and device based on group signature, and a federated learning method. Background Art
[0002] With the rapid development of computing technology, machine learning has become a key technology for data analysis and pattern recognition. Machine learning enables computer systems to learn and improve themselves from data, by training algorithms to recognize patterns and features, so that they can make effective predictions or decisions on new data. These algorithms range from simple linear regression (used to predict or estimate the relationship between things) to complex deep learning networks that simulate the way the human brain processes information. Deep learning networks are particularly suitable for tasks such as image and speech recognition, and their importance is becoming more and more prominent as fields such as medicine, finance, and social media become increasingly dependent on large data sets.
[0003] Taking the medical field as an example, machine learning technology has been widely used in disease diagnosis, disease progression prediction, and personalized treatment plan formulation. These applications rely on processing sensitive patient data, such as medical records, biometric information, and gene sequences. In personal application areas, such as smartphone applications and social network services, machine learning technology also plays an important role in personalized recommendations and user behavior analysis, but the data processed often involves users' personal preferences and behavioral habits.
[0004] In this context, data privacy and confidentiality issues are particularly important. Although data provides valuable insights, if not handled properly, it may lead to privacy breaches and even legal and ethical issues. For example, the leakage of medical data not only violates the privacy rights of patients, but may also lead to social distrust and legal proceedings. In terms of personal applications, improper use of user data may trigger a crisis of trust and affect the company's brand image.
[0005] Faced with this situation, although traditional centralized machine learning has advantages in computing efficiency, the privacy protection risks it faces when processing sensitive data cannot be ignored. Traditional machine learning methods usually rely on centralizing large amounts of data to a single data center or server for processing. Although this centralized approach allows algorithms to access large data sets, thereby improving learning efficiency and accuracy, it also has obvious flaws: on the one hand, centralized storage of large amounts of sensitive data increases the risk of theft or abuse; on the other hand, the transmission of all data to the central server not only consumes a lot of bandwidth, but may also expose privacy information during the transmission process (for example, IP addresses). This risk is particularly prominent in the current environment of frequent cyber attacks and data security incidents.
[0006] To address these issues, federated learning, an emerging machine learning method, has emerged. Federated learning no longer relies on centralized storage of data, but instead keeps the data on the original data source (such as a user's smartphone or a hospital's server). In this model, machine learning models are sent to various data sources, learn independently on these local devices, and then only the learned model updates (not the original data) are transmitted back to the central server. The central server improves the global model by synthesizing these updates. The advantage of this approach is that the original data does not need to leave its source, greatly reducing the risk of data leakage and also reducing the demand for bandwidth.
[0007] However, although federated learning has made some progress in improving content privacy, it still faces challenges in hiding the address of the data source. For example, in the medical field, even if the patient's specific health data is protected, if the data can be identified as coming from a specific hospital or region, it may still expose the patient's identity or health status. In the field of personal applications, even if the user's specific behavioral data is anonymized, the leakage of the geographic location information of the data source may reveal the user's personal habits and behavior patterns.
[0008] Therefore, it is particularly urgent to develop a federated learning method that can simultaneously protect data content and hide the data source address. Summary of the invention
[0009] In view of the deficiencies in the prior art, the present invention provides a data transmission method, program, medium and device based on group signature, and a federated learning method, which solves the problem of data source address exposure in federated learning in the prior art.
[0010] In the first aspect, according to an embodiment of the present invention, a data transmission method based on group signature is provided, which is used for a local node in federated learning to upload its trained model update parameters to an aggregation server, and the method comprises the following steps: S1, using a group certificate and a node private key to sign the model update parameters to obtain a node signature; S2, sending the group certificate, the model update parameters and the node signature to the aggregation server; wherein, a plurality of local nodes participating in federated learning together constitute a group, the group signature group generated by the group includes a group private key, the node private key is one of the group private keys, and the group certificate is generated based on the node private key.
[0011] According to an embodiment of the present invention, the group signature group is generated based on random parameters.
[0012] According to an embodiment of the present invention, the group certificate is generated according to the following steps: P1, the local node generates a first intermediate parameter based on its node private key and passes it to the group administrator in the group; P2, the group administrator obtains the group private key based on the first intermediate parameter, generates a group certificate based on the group private key and random parameters, and returns the group certificate to the local node.
[0013] According to an embodiment of the present invention, a differential privacy method is used in the process of training local nodes to obtain model update parameters.
[0014] In a second aspect, according to another embodiment of the present invention, a federated learning method is provided, in which a local node trains a network model received from an aggregation server based on its local data to obtain model update parameters, and uploads the model update parameters to the aggregation server for aggregation, and the local node adopts a group signature-based data transmission method as described in the first aspect to transmit the model update parameters to the aggregation server.
[0015] In a third aspect, according to another embodiment of the present invention, a computer program product is provided, wherein the computer program product includes a computer program code, and when the computer program code runs on an electronic device, the electronic device executes a data transmission method based on a group signature as described in the first aspect.
[0016] In a fourth aspect, according to another embodiment of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, and the computer program can be executed by a processor to implement the steps of a group signature-based data transmission method described in the first aspect.
[0017] In the fifth aspect, according to another embodiment of the present invention, an electronic device is provided, comprising: one or more processors; and a memory, wherein the memory is used to store executable instructions; the one or more processors are configured to implement the steps of the group signature-based data transmission method described in the first aspect by executing the executable instructions.
[0018] The technical principle of the present invention is: in federated learning, local nodes form a group and generate a group signature group, which includes a group private key; for each local node, a group certificate is generated based on the group private key and the node private key to sign the model parameters generated by each local node training model to obtain a node signature; after sending the group certificate, group signature and model parameters to the aggregation server, the aggregation server aggregates the model parameters by verifying the group signature.
[0019] Compared with the prior art, the present invention has the following beneficial effects: by adopting the above-mentioned model parameter transmission method based on group signature, it solves the technical problem of exposing the data source address in the existing data transmission, and produces the technical effect of hiding and protecting the data source address. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] Figure 1 The figure is a flow chart of a data transmission method based on group signature according to an embodiment of the present invention. DETAILED DESCRIPTION
[0021] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.
[0022] As described in the background technology, there is a problem of data source address leakage when the local node uploads the model parameters in the federated learning of the prior art. In view of the above problem, according to an embodiment of the present invention, a data transmission method based on group signature is provided, which is used for the local node in federated learning to upload the model update parameters obtained by its training to the aggregation server. The method includes the following steps: S1, using the group certificate and the node private key to sign the model update parameters to obtain the node signature; S2, sending the group certificate, the model update parameters and the node signature to the aggregation server; wherein, a plurality of local nodes participating in the federated learning together constitute a group, the group signature group generated by the group includes a group private key, the node private key is one of the group private keys, and the group certificate is generated based on the node private key. By using the group signature, the method can more effectively protect the privacy of local nodes (e.g., individuals and devices). In this method, even if the data is used for a common learning purpose, the data source information (such as identity and geographic location) of the local node (such as an individual) will not be leaked to other participants or third parties.
[0023] According to an embodiment of the present invention, the group signature group is generated based on random parameters. In detail, the group administrator in the group randomly selects some basic parameters to generate a group signature group, wherein the group signature group includes a group public key and a group private key. The group public key is published to the network so that everyone can query it. However, the group private key is only kept by the group administrator to ensure that no one else can obtain the group private key.
[0024] According to an embodiment of the present invention, the group certificate is generated according to the following steps: P1, the local node generates a first intermediate parameter based on its node private key and transmits it to the group administrator in the group; P2, the group administrator obtains the group private key based on the first intermediate parameter, generates a group certificate based on the group private key and a random parameter, and returns the group certificate to the local node. In detail, the local node randomly selects a node private key, and then generates a first intermediate parameter from the node private key. The first intermediate parameter corresponds to the node private key one by one, but the node private key cannot be deduced from the first intermediate parameter. The local node transmits the first intermediate parameter to the group administrator. After the group administrator receives the first intermediate parameter, the group private key is taken out, and the random parameter is selected to form a group certificate for this local node, and the group certificate is returned to the local node. Group certificates are used for signing and are different from each other. In addition, for the newly added local node, the steps of generating a group certificate are similar and will not be repeated. Although the group certificate and the node private key correspond one to one, it is a mathematically difficult problem to reverse the node private key from the group certificate, thereby ensuring that even the group administrator cannot impersonate the local node to sign.
[0025] In addition, in order to further ensure the privacy of the data source, according to an embodiment of the present invention, a differential privacy method is used in the process of training the local node to obtain the model update parameters. As we all know, differential privacy is a technology designed to provide strong data privacy protection. Its core principle is to add a certain amount of random noise to the published data or query results, so that the presence or absence of a single data record has a negligible impact on the results. In this way, even if the attacker has all the data except the target record, they cannot determine whether the target record exists in the data set.
[0026] In addition, according to an embodiment of the present invention, the method can also trace the update parameters. Federated learning usually requires the use of massive data from many nodes, so abnormal situations are inevitable. If the wrong update parameters are released due to unintentional or malicious operations, the group administrator can open the signature under specific conditions (such as multiple incorrect releases, member requests, etc.) to confirm the identity of the group member who signed the signature, and take corresponding measures (such as warnings, removing members, etc.) to maintain the overall reputation of the group.
[0027] By using the above-mentioned group signature, the data transmission method according to an embodiment of the present invention can verify the integrity of the data during transmission and processing. This ensures that the data involved in the learning has not been tampered with, and increases the reliability of the model results. In addition, this method allows new participants to easily join the group and can adapt to federated learning projects of different sizes and types, from small professional groups to large cross-domain applications. In addition, since the privacy of the data source is better protected, participants may be more willing to share their data for federated learning, thereby increasing the diversity and richness of the data, which is beneficial to improving the performance of the overall model.
[0028] In a second aspect, according to another embodiment of the present invention, a federated learning method is provided, in which a local node trains a network model received from an aggregation server based on its local data to obtain model update parameters, and uploads the model update parameters to the aggregation server for aggregation, wherein the local node adopts a group signature-based data transmission method as described above to transmit the model update parameters to the aggregation server.
[0029] 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 of the present invention, which should be included in the scope of the claims of the present invention.
Claims
1. A data transmission method based on group signature, which is used for local nodes in federated learning to upload the model update parameters obtained by their training to the aggregation server. It is characterized in that The method comprises the following steps: S1. Use the group certificate and node private key to sign the model update parameters to obtain the node signature; S2. Sending the group certificate, the model update parameter and the node signature to an aggregation server; Wherein, a plurality of local nodes participating in federated learning together constitute a group, a group signature group generated by the group includes a group private key, the node private key is one of the group private keys, and the group certificate is generated based on the node private key; The group certificate is generated according to the following steps: P1, the local node generates a first intermediate parameter based on its node private key and transmits it to the group administrator in the group; P2. The group administrator obtains a group private key based on the first intermediate parameter, generates a group certificate based on the group private key and a random parameter, and returns the group certificate to the local node.
2. A data transmission method based on group signature as claimed in claim 1, It is characterized in that The group signature group is generated based on random parameters.
3. A data transmission method based on group signature as claimed in claim 2, It is characterized in that The differential privacy method is used in the process of training local nodes to obtain model update parameters.
4. A federated learning method for a local node to train a network model received from an aggregation server based on its local data to obtain model update parameters, and upload the model update parameters to the aggregation server for aggregation, It is characterized in that The local node adopts a group signature-based data transmission method as described in any one of claims 1 to 3 to transmit the model update parameters to the aggregation server.
5. A computer program product, It is characterized in that The computer program product includes a computer program code, and when the computer program code is executed on an electronic device, the electronic device executes a group signature-based data transmission method as claimed in any one of claims 1 to 3.
6. A computer-readable storage medium, It is characterized in that A computer program is stored thereon, and the computer program can be executed by a processor to implement the steps of a data transmission method based on a group signature as described in any one of claims 1 to 3.
7. An electronic device, It is characterized in that include: one or more processors; as well as A memory, wherein the memory is used to store executable instructions; The one or more processors are configured to implement the steps of the group signature-based data delivery method according to any one of claims 1 to 3 by executing the executable instructions.
Citation Information
Patent Citations
Federated learning method based on trusted execution environment
CN111241580A
Federal learning privacy protection method based on improved group signature
CN118133311A
