Security auditing method, device and system for federal learning execution process

Abstract

The invention relates to a security auditing method, device and system for a federal learning execution process and a storage medium, and the method comprises the steps: obtaining transmission data received and / or sent by a participant, and extracting control flow information and / or algorithm flow information from the transmission data; performing security auditing on the execution process of the federated learning task according to the control flow information; and / or performing security auditing on the execution process of the federated learning algorithm according to the algorithm flow information; and generating an auditing result under the condition that an exception is found in the security auditing process or the security auditing is ended. Through the security auditing method provided by the invention, the orderly execution and security of the execution process under the condition that the federated learning task particularly relates to a plurality of participants are ensured, and the process attacks on different levels of federated learning are effectively defended.

Description

technical field [0001] The present invention relates to the technical field of data security and privacy protection, in particular to a security audit method, device, system and storage medium of a federated learning execution process. Background technique [0002] With the development of application fields such as artificial intelligence and big data mining analysis, the demand for data volume is increasing. High-quality data often comes from application data generated and accumulated in business activities, and application data is often scattered in the hands of different organizations and individuals, and cross-industry and cross-field application data is often scattered. With the importance of data ownership, user privacy, and data security receiving more attention, and as laws and regulations impose stricter constraints and requirements on data collection and processing, organizations or individuals who control application data are often unwilling or not suitable means...

AI Technical Summary

Problems solved by technology

[0003] The federated learning algorithm theoretically guarantees the security of data privacy, but in the actual industrial landing, there are still many security risks that need to be considered, such as different failures or communication interferences that may be encountered during the federated learning task , there may even be malicious attacks by malicious participants or malicious processes aimed at destroying the federated learning task, which greatly reduces the execution efficiency of the federated learning task, especially when multiple participants are involved in collaborative model training and application In this context, how to defend against malicious attacks aimed at destroying federated learning tasks and ensure the execution efficiency of federated learning tasks is particularly important, and it is also a technical problem that needs to be solved urgently at the industrial level

Images