Training a machine learning model in a distributed privacy-preserving environment

A computer program product for labeling and utilizing unlabeled public data, the computer program product comprising a non-transitory computer readable storage medium having program instructions embodied therewith, the program instructions readable and executable by a computer to:

• train one or more local machine learning models with local private data to create one or more trained models, wherein the one or more trained models are digital descriptions of an anomalous event that has occurred to a certain piece of equipment, and wherein the local private data is hidden from a public global model;
• generate a label for each of the local private data using the one or more trained models, wherein each label describes the corresponding local private data;
• apply the label for each of the local private data to unlabeled public data to create labeled public data, wherein the unlabeled public data is sensor data from a device from an Internet of Things (IoTs), wherein the device is equipment whose identity is secret, and wherein the sensor data is a data vector of raw data that describes an operational state of the device from the IoTs;
• input the data vector to the public global model to generate an output that the equipment is malfunctioning;
• compare a first quantity of bits in a local data vector from the local private data to a second quantity of bits in a public data vector from the unlabeled public data;
• determine that the first quantity of bits matches the second quantity of bits;
• in response to the first quantity of bits matching the second quantity of bits, assign a label for the local private data to the unlabeled public data to create the labeled public data;
• input the labeled public data into the public global model that uses public data;
• and
• train the public global model by using the labeled public data as input to the public global model.

A system comprising:

• one or more processors;
• one or more computer readable memories operably coupled to the one or more processors;
• one or more computer readable storage mediums operably coupled to the one or more computer readable memories;
• and
• program instructions stored on at least one of the one or more computer readable storage mediums for execution by at least one of the one or more processors via at least one of the one or more computer readable memories, the program instructions comprising:
• program instructions to train one or more local machine learning models with local private data to create one or more trained models, wherein the one or more trained models are digital descriptions of an anomalous event that has occurred to a certain piece of equipment, and wherein the local private data is hidden from a public global model;
• program instructions to generate a label for each of the local private data using the one or more trained models, wherein each label describes the corresponding local private data;
• program instructions to apply the label for each of the local private data to unlabeled public data to create labeled public data, wherein the unlabeled public data is sensor data from a device from an Internet of Things (IoTs), wherein the device is equipment whose identity is secret, and wherein the sensor data is a data vector of raw data that describes an operational state of the device from the IoTs;
• program instructions to input the data vector to the public global model to generate an output that the equipment is malfunctioning;
• program instructions to compare a first quantity of bits in a local data vector from the local private data to a second quantity of bits in a public data vector from the unlabeled public data;
• program instructions to determine that the first quantity of bits matches the second quantity of bits;
• and
• program instructions to, in response to the first quantity of bits matching the second quantity of bits, assign a label for the local private data to the unlabeled public data to the create the labeled public data;
• program instructions to input the labeled public data into the public global model that uses public data;
• and
• program instructions to train the public global model by using the labeled public data as input to the public global model.