树模型训练方法、装置和系统
By calculating the gain of encrypted label distribution information and segmentation strategy, and utilizing homomorphic encryption and public key synthesis techniques, the problem of label data leakage during tree model training in vertical federated learning is solved, thus achieving a secure and efficient tree model training process.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HUAWEI TECH CO LTD
- Filing Date
- 2020-12-31
- Publication Date
- 2026-07-17
AI Technical Summary
In vertical federated learning, there is a security risk of labeled data leakage when the unlabeled party obtains the distribution of samples in each node of the tree model during the construction process.
By calculating the gain of encrypted tag distribution information and segmentation strategies, and utilizing homomorphic encryption and public key synthesis techniques, we ensure that unlabeled and labeled parties exchange and calculate intermediate parameters in ciphertext, avoiding the direct transmission of plaintext data and reducing the risk of using sample set distribution to infer tag data.
It improves the security of vertical federated learning, prevents the leakage of labeled data, ensures data privacy protection, and enables an efficient tree model training process.
Smart Images
Figure CN114692717B_ABST