Protected audience selection

A non-transitory computer-readable storage medium storing processor-executable computer program instructions that, when executed, cause a computer processor to perform a method, the method comprising:

• by a modeling module:
• receiving a respective history of each of a plurality of browsers, each history comprising a first plurality of media consumption events and a corresponding label;
• generating a privatized modeling history corresponding to each respective history of each of the plurality of browsers by removing each of the corresponding labels;
• determining a first frequency of each of a second plurality of media consumption events among a first plurality of privatized modeling histories and a second frequency of each of the second plurality of media consumption events among a second plurality of privatized modeling histories, wherein each of the first plurality of privatized modeling histories comprise a specific conversion event and wherein each of the second plurality of privatized modeling histories do not comprise the specific conversion event;
• building a model based on the first frequency and the second frequency;
• and
• sending the model to a targeting module;
• by the targeting module:
• receiving, from the modeling module, the model;
• receiving, from a real-time bidding exchange, a history of a targetable browser comprising a third plurality of media consumption events;
• and
• responsive to receiving the history of a targetable browser:
• generating a privatized targeting history of the targetable browser by removing a specific media consumption event from the history of the targetable browser;
• applying the model to the privatized targeting history of the targetable browser;
• and
• sending a response to the real-time bidding exchange according to a result of applying the model.

A system comprising:

• a processor;
• and
• a non-transitory computer-readable storage medium storing processor-executable computer program instructions that, when executed, cause a computer processor to perform a method, the method comprising:
• by a modeling module:
• receiving a respective history of each of a plurality of browsers, each history comprising a first plurality of media consumption events and a corresponding label;
• generating a privatized modeling history corresponding to each respective history of each of the plurality of browsers by removing each of the corresponding labels;
• determining a first frequency of each of a second plurality of media consumption events among a first plurality of privatized modeling histories and a second frequency of each of the second plurality of media consumption events among a second plurality of privatized modeling histories, wherein each of the first plurality of privatized modeling histories comprise a specific conversion event and wherein each of the second plurality of privatized modeling histories do not comprise the specific conversion event;
• building a model based on the first frequency and the second frequency;
• and
• sending the model to a targeting module;
• by the targeting module:
• receiving, from the modeling module, the model;
• receiving, from a real-time bidding exchange, a history of a targetable browser comprising a third plurality of media consumption events;
• and
• responsive to receiving the history of a targetable browser:
• generating a privatized targeting history of the targetable browser by removing a specific media consumption event from the history of the targetable browser;
• applying the model to the privatized targeting history of the targetable browser;
• and
• sending a response to the real-time bidding exchange according to a result of applying the model.