INTEGRATING DISTRIBUTED MACHINE LEARNING ALGORITHM AND DIFFERENTIAL PRIVACY MECHANISM INTO THE CROWDSENSING SYSTEM

Portable intelligent devices such as mobile phones with built-in sensors and Internet access have become the basis of all intelligent personal gadgets. A large number of devices have the ability to collectively collect and perform data processing on an unprecedented scale. In this paper, a software that preserves the confidentiality of machine learning for a group of smartphones is presented, which allows solving a wide range of problems related to machine learning of a group of devices with differential privacy conditions. The system provides the ability to teach classifiers or forecasting programs online, on cursor data, privately and with minimal computing costs on devices and servers.