Enhanced clean-room based machine learning

A method for encapsulating machine learning in a clean room to protect data used in the machine learning, the method comprising:

• retrieving, by at least one processor operating within a clean room restricting access to inputs to and outputs of machine learning models encapsulated within the clean room, from data storage in the clean room, data shared between multiple parties for use in machine learning to generate a goal-based output;
• training, by the at least one processor, a machine learning model of the machine learning models encapsulated within the clean room based on training data as inputs to the machine learning model to generate the goal-based output;
• selecting, by the at least one processor, the machine learning model based on a goal provided by at least one of the multiple parties;
• generating, by the at least one processor, using the data as inputs to the selected machine learning model in the clean room, a first set of probabilities encapsulated within the clean room, wherein each probability of the first set of probabilities is a first probability that a respective user may perform an action;
• generating, by the at least one processor, using the selected machine learning model encapsulated within the clean room and the first set of probabilities, a second set of probabilities encapsulated within the clean room, wherein each probability of the second set of probabilities is a second probability that a respective user may perform the action;
• generating, by the at least one processor, in the clean room, based on the second set of probabilities, the goal-based output, wherein the goal-based output is encapsulated within the clean room;
• identifying, by the at least one processor, an egress location for the goal-based output, wherein at least one of the multiple parties provides the egress location;
• and
• sending, by the at least one processor, the goal-based output from the clean room to the egress location.

A system for encapsulating machine learning in a clean room to protect data used in the machine learning, the system comprising at least one processor coupled to memory, the at least one processor operating in a clean room and configured to:

• retrieve, from data storage in a clean room restricting access to inputs to and outputs of machine learning models encapsulated within the clean room, data shared between multiple parties for use in machine learning to generate a goal-based output;
• train a machine learning model of the machine learning models based on training data as inputs to the machine learning model to generate the goal-based output;
• select the machine learning model based on a goal indicated by at least one of the multiple parties;
• generate, using the data as inputs to the selected machine learning model in the clean room, a first set of probabilities encapsulated within the clean room, wherein each probability of the first set of probabilities is a first probability that a respective user may perform an action;
• generate, using the selected machine learning model encapsulated within the clean room and the first set of probabilities, a second set of probabilities encapsulated within the clean room, wherein each probability of the second set of probabilities is a second probability that a respective user may perform the action;
• generate, based on the second set of probabilities, in the clean room, the goal-based output, wherein the goal-based output is encapsulated within the clean room;
• identify an egress location for the goal-based output, wherein at least one of the multiple parties provides the egress location;
• and
• send the goal-based output from the clean room to the egress location.