Encrypted Viewer Preference Signals for Privacy-Preserving Document Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Content aggregators often present users with document summaries that are not narrowly tailored to their interests and fail to protect user privacy, monetizing reader interests by using them for targeted advertising.
Innovation Solution
A method that detects document attribute preferences by computing attribute scores based on viewer interactions, selecting and arranging documents accordingly, and maintaining user privacy by encrypting interaction data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If content aggregators use identified viewer interests to provide more-directed advertising, then advertising effectiveness is improved, but viewer privacy is compromised
Solution Approach 1:
The patent introduces encrypted interest signals as an intermediary between viewer behavior and advertising delivery. The encryption system allows advertising effectiveness to be maintained through decrypted signals while preserving viewer privacy by ensuring the encryption keys never leave the viewer's device. This resolves the contradiction by enabling targeted advertising without direct access to raw interest data.
Solution Approach 2:
The patent creates encrypted copies of viewer interest signals that can be transmitted to advertisers without revealing the actual interest data. These cryptographic representations preserve the utility of interest information for advertising while preventing privacy compromise, as the original unencrypted data never leaves the viewer's device.
2Adaptability or versatility
If content aggregators present a broad selection of document summaries, then content diversity is improved, but relevance to viewer interests deteriorates
Solution Approach 1:
The patent implements dynamic document selection that adapts to changing viewer interests over time. The system continuously monitors encrypted interest signals and adjusts document recommendations accordingly, allowing the content selection to be both diverse (adaptable to various interests) and relevant (tailored to current interests). This resolves the contradiction by making the selection process dynamic rather than static.
Solution Approach 2:
The patent changes the parameter of interest representation from raw data to encrypted signals, allowing the system to maintain both diversity and relevance. By operating with encrypted interest parameters rather than explicit interest profiles, the system can explore diverse content while still measuring and responding to viewer preferences with high precision.
3Measurement precision
If content aggregators collect detailed viewer interaction data, then personalization accuracy is improved, but data security requirements increase
Solution Approach 1:
The patent inverts the traditional approach by not collecting raw interaction data for personalization, but rather having the viewer's device generate encrypted interest signals from interaction data locally. This inversion maintains personalization accuracy through precise encrypted signals while eliminating the security burden of storing and protecting large volumes of raw interaction data on server systems.
Data Source
AI summary
Some embodiments of the invention provide a method that detects document attribute preferences for viewing documents (e.g., articles) on a device. The method displays several documents on a display screen of the device, with each document having one or more associated attributes. In some embodiments, different documents can be associated with the same attribute or the same number of attributes, or can be associated with different attributes or a different number of attributes. In some embodiments, the method detects viewer interactions with one or more displayed documents, and based on these interactions, computes attribute scores for several attributes associated with a number of the documents. The computed attribute scores identify a preference ranking for attributes associated with the documents. Some embodiments use the computed attribute scores to select subsequent documents for display on the device and/or to arrange the display of subsequent documents on the device.