Split Model Inference for Secure Client-Server Data Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
User data security is compromised during inference processes on servers and clients in neural network models.
Innovation Solution
A model inference method where a client splits processing results and sends partial data to a server, using partial model parameters and encryption algorithms to ensure data security and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If the server obtains all user data for inference, then the inference completeness is improved, but the data security deteriorates
Solution Approach 1:
The user data is segmented into multiple parts (first data and second data) where the first data remains on the client and only the second data is transmitted to the server. This segmentation allows the server to perform inference on partial data without accessing complete user data, thus maintaining both inference completeness and data security.
Solution Approach 2:
The essential information needed for inference is extracted and transmitted to the server, while the remaining data that is not critical for the inference task is retained on the client side. This extraction principle allows the server to obtain sufficient information for accurate inference without receiving complete sensitive data.
2Measurement precision
If the client sends all user data to the server, then the inference accuracy is improved, but the bandwidth consumption increases
Solution Approach 1:
By segmenting user data and transmitting only the necessary portion (second data) to the server, the bandwidth consumption is significantly reduced while maintaining sufficient inference accuracy. The server receives minimal data required for the inference task without unnecessary additional information.
3Object-affected harmful factors
If the client processes all data locally, then the data security is improved, but the computational burden on the client increases
Solution Approach 1:
The computational task is segmented between client and server: the client performs local processing on first data and the server processes second data. This division reduces the computational burden on the client while maintaining data security, as the client only needs to process a portion of the data locally.
Solution Approach 2:
The server acts as an intermediary that handles the computational processing of transmitted data, relieving the client of the burden of processing all data locally. The client sends minimal data to the server which then performs the computationally intensive inference tasks.
Data Source
AI summary
A model inference method and apparatus are disclosed, and relates to the field of machine learning technologies. A client and a server use respective deployed models to process different parts of user data, to obtain respective output results. In addition, the client obtains the output result of the server, and obtains an inference result based on the output results of the server and the client. Compared with a case in which the server needs to obtain all the user data in an inference process, in this application, the server obtains only a part of the user data. As the server cannot obtain, based on the part of the user data, all content included in the user data, security of the user data is ensured.