Differentially Private Frequency Deduplication

A method for deduplicating data sets corresponding to frequencies of user interaction with a plurality of providers, the method comprising:

• receiving, by a data processing system comprising one or more processors and one or more memories, a first set of two or more vector data structures from a first provider of the plurality of providers, each of the first set of two or more vector data structures encoding, with differential privacy, user identifiers associated with a number of interactions with the first provider, having a target interaction type, that matches a respective frequency of each of the first set of two or more vector data structures;
• receiving, by the data processing system, a second set of two or more vector data structures from a second provider of the plurality of providers, each of the second set of two or more vector data structures encoding, with differential privacy, user identifiers associated with a number of interactions with the second provider, having the target interaction type, that matches a respective frequency of each of the second set of two or more vector data structures;
• and
• combining, by the data processing system, the first set of two or more vector data structures and the second set of two or more vector data structures into a deduplicated frequency distribution that defines a total frequency of user interactions having the target interaction type across both the first provider and the second provider for each of a plurality of frequencies.

A system for reducing data sets received from a plurality of providers, the system comprising:

• a network interface configured to communicate with a plurality of servers associated with the plurality of providers;
• and
• one or more processors coupled to the network interface and a memory, the memory comprising instructions stored thereon that, when executed by the one or more processors, cause the one or more processors to:
• receive, via the network interface, a first set of differentially private sketches from a first provider of the plurality of providers, each of the first set of differentially private sketches encoding, with differential privacy, user identifiers associated with a number of interactions with the first provider, having a target interaction type, that matches a respective frequency of each of the first set of differentially private sketches;
• receive, via the network interface, a second set of differentially private sketches from a second provider of the plurality of providers, each of the second set of differentially private sketches encoding, with differential privacy, user identifiers associated with a number of interactions with the second provider, having the target interaction type, that matches a respective frequency of each of the second set of differentially private sketches;
• generate a combined distribution using the first set of differentially private sketches and the second set of differentially private sketches, the combined distribution deduplicating the data sets represented by the first set of differentially private sketches and by the second set of differentially private sketches and defining a total frequency of user interactions having the target interaction type across both the first provider and the second provider for each of a plurality of frequencies;
• and
• store, in a database, the combined distribution.