Methods and apparatuses for defending against data poisoning attacks in recommender systems

A method comprising:

• obtaining customer interaction data characterizing customer interactions of a customer with an ecommerce marketplace;
• determining an item recommendation based on the customer interaction data using a trained differentially private recommendation model, wherein the trained differentially private recommendation model is more likely to determine the same item recommendation after poisoned data is injected into the customer interaction data than a recommendation model that is not privately trained;
• and
• sending the item recommendation to the customer.

A non-transitory computer readable medium having instructions stored thereon, wherein the instructions, when executed by at least one processor, cause a device to perform operations comprising:

• obtaining customer interaction data characterizing customer interactions of a customer with an ecommerce marketplace;
• determining an item recommendation based on the customer interaction data using a trained differentially private recommendation model, wherein the trained differentially private recommendation model is more likely to determine the same item recommendation after poisoned data is injected into the customer interaction data than a recommendation model that is not privately trained;
• and
• sending the item recommendation to the customer.