一种安全检测方法、装置、设备及可读存储介质
By splitting the machine learning model into front-end and back-end sub-models and performing security checks, the problem of data leakage during model use was solved, thus protecting privacy information.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ALIPAY (HANGZHOU) INFORMATION TECH CO LTD
- Filing Date
- 2022-09-07
- Publication Date
- 2026-07-17
AI Technical Summary
Machine learning models pose a security risk of data leakage during use. Attackers may steal intermediate results from the model to reconstruct the input or output, leading to the leakage of private information.
By splitting the machine learning model into a front-end sub-model and a back-end sub-model, deploying them on different devices, and employing security detection methods, the privacy leakage risks of the front-end sub-model and the back-end sub-model are detected by a first detection model and a second detection model, respectively, thus quantifying the privacy leakage risks of the model.
It enables the quantification of privacy leakage risks of machine learning models, enhances the security of privacy information during model use, and prevents privacy information leakage.
Smart Images
Figure CN116150645B_ABST