# US 12,001,569 B2 · Model training method, model using method, system, trusted node and device > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,001,569 B2 - Document ID: US12001569B2 - Status: Granted - Kind: B2 - Application number: 17/669,922 - Filed: 2022-02-11 - Published: 2022-11-03 - Granted: 2024-06-04 - Earliest claimed: 2021-04-29 - Assignees: Jiangsu Superfluidity Information Technology Co., Ltd - Applicants: Jiangsu Superfluidity Information Technology Co., Ltd - Inventors: Chuan Lv; Xiaoxing Zhang - Web page: https://patentgrep.com/patent/US12001569B2 ## CPC - G06F 21/602: {Providing cryptographic facilities or services} - G06N 10/60: Quantum algorithms, e.g. based on quantum optimisation, quantum Fourier or Hadamard transforms - G06N 20/00: Machine learning - H04L 9/0852: {Quantum cryptography (transmission systems employing electromagnetic waves other than radio waves, e.g. light, infrared H04B10/00; wavelength-division multiplex systems H04J14/02; WDM arrangements H04J14/03)} - H04L 9/0866: {involving user or device identifiers, e.g. serial number, physical or biometrical information, DNA, hand-signature or measurable physical characteristics} - H04L 9/0894: {Escrow, recovery or storing of secret information, e.g. secret key escrow or cryptographic key storage} - H04L 9/16: the keys or algorithms being changed during operation - H04L 63/0428: {wherein the data content is protected, e.g. by encrypting or encapsulating the payload} - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M) ## Abstract Aspects relating to machine learning includes receiving, by a first trusted node, a first target data set sent by a first participant, wherein the first target data set is obtained via encrypting, by the first participant, a data set provided by the first participant based on a first preset encryption mode; decrypting the first target data set, determining first training data, and performing model training for a preset machine learning model based on the first training data to obtain a first intermediate training result; acquiring an encrypted second intermediate training result sent by at least one second trusted node; and performing federated learning for the preset machine learning model based on at least the first intermediate training result and the decrypted second intermediate training result, to update model parameters of the preset machine learning model and obtain a learning completed target model. ## Claims ### Claim 1 A model training method based on federated learning, comprising: receiving, by a first trusted node, a first target data set sent by a first participant, wherein the first target data set is obtained via encrypting, by the first participant, a data set provided by the first participant based on a first preset encryption mode; decrypting the first target data set, determining first training data, and performing model training for a preset machine learning model based on the first training data to obtain a first intermediate training result; acquiring an encrypted second intermediate training result sent by at least one second trusted node; wherein the second intermediate training result is obtained via performing, by the second trusted node, model training for the preset machine learning model based on at least second training data, the second training data is determined after decrypting, by the second trusted node, an encrypted second target data set, and the second target data set is provided by a second participant for federated learning with the first target data set; performing federated learning for the preset machine learning model based on at least the first intermediate training result and the decrypted second intermediate training result, to update model parameters of the preset machine learning model and obtain a learning completed target model, and decrypting the second intermediate training result based on a second quantum key generated by a first quantum key generation device contained in the first trusted node, wherein the second quantum key and a quantum key generated by a second quantum key generation device contained in the second trusted node form a quantum key pair, and the second trusted node encrypts the second intermediate training result based on the quantum key forming the quantum key pair with the second quantum key. ### Claim 2 (depends on 1) The method according to claim 1, further comprising: encrypting the first intermediate training result based on a second preset encryption mode, and sending the encrypted first intermediate training result to the second trusted node; wherein the second trusted node is capable of decrypting the encrypted first intermediate training result to perform model training for the preset machine learning model and update the model parameters of the preset machine learning model. ### Claim 3 (depends on 2) The method according to claim 2, wherein encrypting the first intermediate training result based on the second preset encryption mode comprises: encrypting the first intermediate training result based on a first quantum key generated by a first quantum key generation device contained in the first trusted node. ### Claim 4 (depends on 3) The method according to claim 3, wherein the first quantum key generation device stores a key pool, and the key pool comprises at least one quantum key; and encrypting the first intermediate training result based on the first quantum key generated by the first quantum key generation device contained in the first trusted node comprises: acquiring the first quantum key from the key pool based on a first router contained in the first trusted node, and encrypting the first intermediate training result based on the first quantum key. ### Claim 5 (depends on 1) The method according to claim 1, wherein the first trusted node obtains target parameter values for the model parameters of the learning completed target model to obtain the target model; or the first trusted node obtains target parameter values for a part of model parameters of the learning completed target model to obtain a part of the target model. ### Claim 6 (depends on 1) The method according to claim 1, wherein determining the first training data comprises: performing data identification matching for the first target data set and the second target data set performing federated learning with the first target data set to obtain a data set whose matching result meets a preset rule; and taking the data set in the first target data set whose matching result meets the preset rule as the first training data. ### Claim 7 (depends on 1) The method according to claim 1, wherein the first target data set is obtained via encrypting the data set provided by the first participant based on a third quantum key. ### Claim 8 (depends on 7) The method according to claim 7, wherein the third quantum key is generated by a third quantum key generation device associated with the first participant, and the third quantum key and a quantum key generated by a first quantum key generation device contained in the first trusted node form a quantum key pair; or the third quantum key is generated by the first quantum key generation device contained in the first trusted node and charged into a mobile quantum key storage medium, so that the first participant is capable of acquiring the third quantum key from the mobile quantum key storage medium. ### Claim 9 (depends on 1) The method according to claim 1, wherein the first trusted node is a node in a preset communication network, and the second trusted node joins the preset communication network after being authenticated with a quantum key, so that the second trusted node is capable of performing federated learning with a trusted node in the preset communication network. ### Claim 10 A first trusted node, comprising: a first receiving circuit, configured to receive a first target data set sent by a first participant, wherein the first target data set is obtained via encrypting, by the first participant, a data set provided by the first participant based on a first preset encryption mode; an intermediate training result determination circuit, configured to decrypt the first target data set, determine first training data, and perform model training for a preset machine learning model based on the first training data to obtain a first intermediate training result; an intermediate training result acquisition circuit, configured to acquire an encrypted second intermediate training result sent by at least one second trusted node; wherein the second intermediate training result is obtained via performing, by the second trusted node, model training for the preset machine learning model based on at least second training data, the second training data is determined after decrypting, by the second trusted node, an encrypted second target data set, and the second target data set is provided by a second participant for federated learning with the first target data set; a federated learning circuit, configured to perform federated learning for the preset machine learning model based on at least the first intermediate training result and the decrypted second intermediate training result, to update model parameters of the preset machine learning model and obtain a learning completed target model, and a decryption circuit, configured to decrypt the second intermediate training result based on a second quantum key generated by a first quantum key generation device contained in the first trusted node, wherein the second quantum key and a quantum key generated by a second quantum key generation device contained in the second trusted node form a quantum key pair, and the second trusted node encrypts the second intermediate training result based on the quantum key forming the quantum key pair with the second quantum key. ### Claim 11 (depends on 10) The node according to claim 10, further comprising: an encryption circuit, configured to encrypt the first intermediate training result based on a second preset encryption mode; and a first sending circuit, configured to send the encrypted first intermediate training result to the second trusted node; wherein the second trusted node is capable of decrypting the encrypted first intermediate training result to perform model training for the preset machine learning model and update the model parameters of the preset machine learning model. ### Claim 12 (depends on 11) The node according to claim 11, wherein the encryption circuit is further configured to encrypt the first intermediate training result based on a first quantum key generated by a first quantum key generation device contained in the first trusted node. ### Claim 13 (depends on 12) The node according to claim 12, wherein the first quantum key generation device stores a key pool, and the key pool comprises at least one quantum key; and the encryption circuit is further configured to acquire the first quantum key from the key pool based on a first router contained in the first trusted node, and encrypt the first intermediate training result based on the first quantum key. ### Claim 14 (depends on 10) The node according to claim 10, wherein the first trusted node obtains target parameter values for the model parameters of the learning completed target model to obtain the target model; or the first trusted node obtains target parameter values for a part of model parameters of the learning completed target model to obtain a part of the target model. ### Claim 15 (depends on 10) The node according to claim 10, wherein the intermediate training result determination circuit is further configured to perform data identification matching for the first target data set and the second target data set performing federated learning with the first target data set to obtain a data set whose matching result meets a preset rule, and take the data set in the first target data set whose matching result meets the preset rule as the first training data. ### Claim 16 (depends on 10) The node according to claim 10, wherein the first target data set is obtained via encrypting the data set provided by the first participant based on a third quantum key. ### Claim 17 (depends on 16) The node according to claim 16, wherein the third quantum key is generated by a third quantum key generation device associated with the first participant, and the third quantum key and a quantum key generated by a first quantum key generation device contained in the first trusted node form a quantum key pair; or the third quantum key is generated by the first quantum key generation device contained in the first trusted node and charged into a mobile quantum key storage medium, so that the first participant is capable of acquiring the third quantum key from the mobile quantum key storage medium. ### Claim 18 (depends on 10) The node according to claim 10, wherein the first trusted node is a node in a preset communication network, and the second trusted node joins the preset communication network after being authenticated with a quantum key, so that the second trusted node is capable of performing federated learning with a trusted node in the preset communication network.