Methods and apparatuses for identifying privacy-sensitive users in recommender systems

Patent metadata

Field	Value
Publication number	US-12026285-B2
Application number	US-202117162295-A
Country	US
Kind code	B2
Filing date	Jan 29, 2021
Priority date	Jan 29, 2021
Publication date	Jul 2, 2024
Grant date	Jul 2, 2024

Abstract

A privacy system includes a computing device configured to obtain user transactional data characterizing at least one transaction of a user on an ecommerce marketplace and to determine a privacy vulnerability score of the user by comparing the transactional data to a user vulnerability distribution. The computing device is also configured to send the privacy vulnerability score to a personalization engine.

First claim

What is claimed is:

1. A system comprising:
a computing device configured to:
train a privacy vulnerability model to generate a privacy vulnerability score, wherein the privacy vulnerability model is trained using a generative adversarial network configured to receive historical user transactions and generated user transactions;
obtain transactional activity data characterizing at least one transaction of a user;
determine the privacy vulnerability score of the user by providing the transactional activity data of the user to the privacy vulnerability model, wherein the user is classified based on the privacy vulnerability score and a user vulnerability distribution, wherein the transactional activity data comprises:
activity sequence data including information about a sequence of actions associate with the user;
contextual data providing context information of the sequence of actions; and
taxonomy data including classification and organization information about items the user has interacted with; and
send the privacy vulnerability score to a personalization engine configured to generate an application protocol interface API to allow access to the privacy vulnerability score by a plurality of personalization engines, wherein at least one of the plurality of personalization engines is configured to automatically personalize an interface provided to the user based on the classification of the user.

2. The system of claim 1 , wherein the user vulnerability distribution is determined during training of the privacy vulnerability model.

3. The system of claim 1 , wherein the privacy vulnerability model is trained using a training method comprising:
obtaining user beacon data characterizing actual customer transaction data on the ecommerce marketplace;
generating user sample data characterizing artificial customer transaction data on the ecommerce marketplace; and
inputting the user beacon data and the user sample data into the privacy vulnerability model to discriminate between the user beacon data and the user sample data.

4. The system of claim 2 , wherein the personalization engine implements at least one privacy preserving measure if the privacy vulnerability score is greater than a privacy vulnerability threshold.

5. A method comprising:
training a privacy vulnerability model to generate a privacy vulnerability score, wherein the privacy vulnerability model is trained using a generative adversarial network configured to receive historical user transactions and generated user transactions;
obtaining transactional activity data characterizing at least one transaction of a user;
determining the privacy vulnerability score of the user by providing the transactional activity data of the user to the privacy vulnerability model, wherein the user is classified based on the privacy vulnerability score and a user vulnerability distribution, wherein the transactional activity data comprises:
activity sequence data including information about a sequence of actions associate with the user;
contextual data providing context information of the sequence of actions; and
taxonomy data including classification and organization information about items the user has interacted with; and

sending the privacy vulnerability score to a personalization engine configured to generate an application protocol interface API to allow access to the privacy vulnerability score by a plurality of personalization engines, wherein at least one of the plurality of personalization engines is configured to automatically personalize an interface provided to the user based on the classification of the user.

6. The method of claim 5 , wherein the user vulnerability distribution is determined during training of the privacy vulnerability model.

7. The method of claim 5 , wherein the privacy vulnerability model is trained using a training method comprising:
obtaining user beacon data characterizing actual customer transaction data on the ecommerce marketplace;
generating user sample data characterizing artificial customer transaction data on the ecommerce marketplace; and
inputting the user beacon data and the user sample data into the privacy vulnerability model to discriminate between the user beacon data and the user sample data.

8. The method of claim 6 , wherein the personalization engine implements at least one privacy preserving measure if the privacy vulnerability score is greater than a privacy vulnerability threshold.

9. 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:
training a privacy vulnerability model to generate a privacy vulnerability score, wherein the privacy vulnerability model is trained using a generative adversarial network configured to receive historical user transactions and generated user transactions;
obtaining transactional activity data characterizing at least one transaction of a user;
determining the privacy vulnerability score of the user by providing the transactional activity data of the user to the privacy vulnerability model, wherein the user is classified based on the privacy vulnerability score and a user vulnerability distribution, wherein the transactional activity data comprises:
activity sequence data including information about a sequence of actions associate with the user;
contextual data providing context information of the sequence of actions; and
taxonomy data including classification and organization information about items the user has interacted with; and

sending the privacy vulnerability score to a personalization engine configured to generate an application protocol interface API to allow access to the privacy vulnerability score by a plurality of personalization engines, wherein at least one of the plurality of personalization engines is configured to automatically personalize an interface provided to the user based on the classification of the user.

10. The non-transitory computer readable medium of claim 9 , wherein the user vulnerability distribution is determined during training of the privacy vulnerability model.

11. The non-transitory computer readable medium of claim 9 , wherein the privacy vulnerability model is trained using a training method comprising:
obtaining user beacon data characterizing actual customer transaction data on the ecommerce marketplace;
generating user sample data characterizing artificial customer transaction data on the ecommerce marketplace; and
inputting the user beacon data and the user sample data into the privacy vulnerability model to discriminate between the user beacon data and the user sample data.

12. The non-transitory computer readable medium of claim 10 , wherein the personalization engine implements at least one privacy preserving measure if the privacy vulnerability score is greater than a privacy vulnerability threshold.

13. The system of claim 1 , wherein the privacy vulnerability model includes a discriminator including a convolutional neural network.

14. The method of claim 5 , wherein the privacy vulnerability model includes a discriminator including a convolutional neural network.

15. The non-transitory computer readable medium of claim 9 , wherein the privacy vulnerability model includes a discriminator including a convolutional neural network.

Assignees

• Walmart Apollo Llc

Inventors

• Achan Kannan
• Sharma Durga Deepthi Singh
• Shahrasbi Behzad
• Agrawal Saurabh
• Mani Venugopal
• Wadhwa Soumya
• Motwani Kamiya
• Korpeoglu Evren
• Kumar Sushant

Classifications

• G06N3/094

  Adversarial learning · CPC title

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/0475

  Generative networks · CPC title

• G06N20/00

  Machine learning · CPC title