Systems and methods to maintain user privacy while providing recommendations
A systematic method of introducing obfuscating “organic” noise to a user's content engagement history leverages a recommender system by creating a public history on a client device which is a superset of the user's true engagement history. The method builds up the superset history over time through a client's interaction with the recommender system by simulating organic growth in a user's actual engagement history. The organic superset prevents an adversary with access to the underlying recommendation model from readily distinguishing between signal and noise in a user's query and obfuscates the user's engagement history with the recommender system.
1. A method for organic noising of user engagement with an item:
detecting, by one or more processors, user engagement with an item via a client device;
storing, by the one or more processors, the item as a cluster in the client device;
augmenting, by the one or more processors, the cluster to include one or more random items;
sending, by the one or more processors, a recommendation request corresponding to the item and the one or more random items in the cluster to a recommendation server;
receiving, by the one or more processors, based on the recommendation request, a plurality of recommendations from the recommendation server comprising at least one real recommendation corresponding to the item; and
outputting, by the one or more processors, the plurality of recommendations on a display of the client device.
2. The method of claim 1 , wherein sending the recommendation request to the recommendation server includes sending non-identifying information.
3. The method of claim 2 , wherein the recommendation request does not include a user identifier, a device identifier corresponding to the client device, an Internet Protocol address corresponding to the client device, or other identifier from which a user of the client device may be identified.
4. The method of claim 1 , wherein each of the plurality of recommendations includes a recommended item, an associated item from the cluster, and a scaled weight of the recommended item.
5. The method of claim 4 , wherein the scaled weight indicates a closeness of the recommended item and the associated item.
6. The method of claim 1 , wherein the cluster includes a plurality of similar items that meet a defined measure of similarity.
7. The method of claim 1 , the method further comprising:
detecting, by the one or more processors, additional user engagement with a second item via the client device; and
storing, by the one or more processors, the second item in the cluster.
8. A computer system for organic noising of user engagement with an item, the computer system comprising:
a memory having processor-readable instructions stored therein; and
one or more processors configured to access the memory and execute the processor-readable instructions, which when executed by the one or more processors configures the one or more processors to perform a plurality of functions, including functions for:
detecting, by the one or more processors, user engagement with an item via a client device;
storing, by the one or more processors, the item as a cluster in the client device;
augmenting, by the one or more processors, the cluster to include one or more random items;
sending, by the one or more processors, a recommendation request corresponding to the item and the one or more random items in the cluster to a recommendation server;
receiving, by the one or more processors, based on the recommendation request, a plurality of recommendations from the recommendation server comprising at least one real recommendation corresponding to the item; and
outputting, by the one or more processors, the plurality of recommendations on a display of the client device.
9. The computer system of claim 8 , wherein sending the recommendation request to the recommendation server includes sending non-identifying information.
10. The computer system of claim 9 , wherein the recommendation request does not include a user identifier, a device identifier corresponding to the client device, an Internet Protocol address corresponding to the client device, or other identifier from which a user of the client device may be identified.
11. The computer system of claim 8 , wherein each of the plurality of recommendations includes a recommended item, an associated item from the cluster, and a scaled weight of the recommended item.
12. The computer system of claim 11 , wherein the scaled weight indicates a closeness of the recommended item and the associated item.
13. The computer system of claim 8 , wherein the cluster includes a plurality of similar items that meet a defined measure of similarity.
14. The computer system of claim 8 , the functions further comprising:
detecting, by the one or more processors, additional user engagement with a second item via the client device; and
storing, by the one or more processors, the second item in the cluster.
15. A non-transitory computer-readable medium containing instructions for organic noising of user engagement with an item, the instructions comprising:
detecting user engagement with an item via a client device;
storing the item as a cluster in the client device;
augmenting the cluster to include one or more random items;
sending a recommendation request corresponding to the item and the one or more random items in the cluster to a recommendation server;
receiving, based on the recommendation request, a plurality of recommendations from the recommendation server comprising at least one real recommendation corresponding to the item; and
outputting the plurality of recommendations on a display of the client device.
16. The non-transitory computer-readable medium of claim 15 , wherein sending the recommendation request to the recommendation server includes sending non-identifying information.
17. The non-transitory computer-readable medium of claim 16 , wherein the recommendation request does not include a user identifier, a device identifier corresponding to the client device, an Internet Protocol address corresponding to the client device, or other identifier from which a user of the client device may be identified.
18. The non-transitory computer-readable medium of claim 15 , wherein each of the plurality of recommendations includes a recommended item, an associated item from the cluster, and a scaled weight of the recommended item.
19. The non-transitory computer-readable medium of claim 18 , wherein the scaled weight indicates a closeness of the recommended item and the associated item.
20. The non-transitory computer-readable medium of claim 15 , wherein the cluster includes a plurality of similar items that meet a defined measure of similarity.