Privacy-Preserving Recommendations Using Contextual Embeddings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing web technologies rely on monitoring and collecting user data through methods like third-party cookies and fingerprinting, which compromise user privacy and require significant network resources, and there is a need for a more privacy-preserving approach to provide relevant digital components without collecting personal data.
Innovation Solution
A method involving user devices generating embedding vectors from contextual data, adding noise to preserve privacy, and transmitting interest group data to a processing server that builds models to predict content categories, allowing for relevant content provision without individual user data collection, using privacy-preserving algorithms to anonymize data exchanges.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If third-party cookies and fingerprinting are used to monitor and collect user data, then relevant digital components can be provided to users, but user privacy is compromised and network resources are consumed
Solution Approach 1:
The patent extracts only the necessary contextual information (URL, digital component requests) from user browsing behavior and eliminates the collection of personal identifiable data. By taking out only the essential features needed for recommendation while discarding personal data, the system achieves relevance without privacy compromise
Solution Approach 2:
The patent introduces an intermediary processing server that acts as a mediator between user devices and content providers. This server receives contextual data, generates embedding vectors, and facilitates privacy-preserving exchanges without directly collecting or storing personal user information, thus protecting privacy while enabling relevant content delivery
2Adaptability or versatility
If third-party cookies and fingerprinting are used to monitor and collect user data, then relevant digital components can be provided to users, but significant network resources are required
Solution Approach 1:
The patent extracts only the essential contextual features (URL, digital component requests) needed for generating embedding vectors, eliminating the need to transmit and process large quantities of personal data across the network. This reduces network resource consumption while maintaining recommendation effectiveness
Solution Approach 2:
The patent transforms raw contextual data into compressed embedding vectors that capture essential patterns and relationships. By changing the data representation from detailed personal information to condensed feature vectors, the system reduces the quantity of data that needs to be transmitted and processed over the network
3Object-affected harmful factors
If embedding vectors and privacy-preserving algorithms are used to anonymize data, then user privacy is enhanced, but data processing complexity increases
Solution Approach 1:
The patent performs preliminary action by pre-computing and storing embedding vectors for common content categories and patterns. This allows the system to quickly generate privacy-preserving anonymized data without complex real-time processing, reducing overall system complexity while maintaining privacy protection
Solution Approach 2:
The patent simplifies complex data processing by transforming detailed contextual information into condensed embedding vectors that capture essential patterns. This parameter transformation reduces the computational complexity of privacy-preserving operations while maintaining effective anonymization and data utility
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on computer-storage media, for privacy preserving digital component provider. In some implementations, a method includes providing, by a user device and during a browsing session of content page at the user device, (1) a request for a digital component and (2) contextual data representing a context within which the content page is provided for display on the user device; obtaining an embedding vector that represents the contextual data as a set of features and the digital component; generating one or more adjusted embedding vectors for a first interest group, wherein the collection includes the embedding vector adjusted by one or more values; and providing the one or more adjusted embedding vectors to a server for generating a model for the first interest group.