阅读说明：本技术 隐私性增强的联合学习系统 (Joint learning system with enhanced privacy ) 是由 阿德里安·弗拉纳根 宽·埃克·谭 付强 于 2019-01-18 设计创作，主要内容包括：一种用户设备,包括：处理器,用于：下载主机器学习模型,其中,所述主机器学习模型用于生成与所述用户设备的应用程序的使用相关的用户推荐；使用所述主机器学习模型和与所述用户设备的用户或与所述用户设备的用户交互中的一个或多个相关的数据来计算所述主机器学习模型的模型更新；使用ε-差分隐私机制对所述计算出的模型更新进行编码；发送ε-差分隐私编码后的模型更新。通过将ε-差分隐私(Differential Privacy,简称DP)应用到从所述用户设备上传到后端服务器的模型更新,增强了联合学习系统的隐私性。因为所述模型更新被散列和随机化,并且不能被单独解码以了解有关所述用户的任何信息,因此进一步增强了所述用户的隐私性。(A user equipment, comprising: a processor to: downloading a host machine learning model, wherein the host machine learning model is used to generate user recommendations related to usage of applications of the user device; computing a model update for the host machine learning model using the host machine learning model and data related to one or more of a user of the user device or a user interaction with the user device; encoding the computed model updates using an epsilon-differential privacy mechanism; and sending the model update after the epsilon-difference privacy coding. By applying Differential Privacy (DP) to model updates uploaded from the user equipment to the back-end server, the Privacy of the joint learning system is enhanced. The privacy of the user is further enhanced because the model updates are hashed and randomized and cannot be decoded separately to learn any information about the user.)