Captured Content Privacy via Local Feature Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for preserving user privacy in captured content often make unacceptable tradeoffs, failing to adequately protect individuals who do not want to be depicted in images or videos without their permission.
Innovation Solution
A privacy setting is enabled on user devices, allowing users to prevent their depiction in captured content by comparing content features with locally stored user features, and if a match is found, the device obfuscates the user's identity in the content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional privacy preservation methods are used, then some level of privacy protection is achieved, but unacceptable tradeoffs are made that compromise content quality or user experience
Solution Approach 1:
The system performs preliminary actions by having users pre-register their visual features (facial recognition data, body characteristics) in advance before any privacy protection is needed. This allows the system to quickly compare captured content against the registered database and apply selective obfuscation only to matching individuals, rather than blanket privacy protection. This resolves the contradiction by enabling precise privacy protection that maintains content usability.
Solution Approach 2:
The patent applies local quality by implementing selective privacy protection that affects only specific portions of the content (individuals matching registered features) while leaving the rest of the content unchanged. This targeted approach ensures that privacy protection is applied locally to affected areas rather than globally to the entire image or video, thereby maintaining overall content quality and usability.
2Reliability
If comprehensive privacy protection is implemented, then user privacy is effectively preserved, but the complexity of the system increases
Solution Approach 1:
The system introduces an intermediary component - a dedicated privacy protection module that acts as a mediator between the content capture function and the final output. This module receives captured content, performs feature extraction, compares against the registered database, and applies obfuscation only when necessary. This intermediary structure manages system complexity by isolating the complex privacy protection logic from the main content processing pipeline.
Solution Approach 2:
The patent uses copying by creating feature representations (copies) of user appearance characteristics that are stored in a database. Instead of storing or processing actual images, the system works with extracted feature data (facial landmarks, body measurements). This copying approach reduces complexity by working with simplified data representations rather than complex image processing throughout the system.
3Reliability
If user features are stored locally on opted-out devices, then user privacy and data security are improved, but the measurement and detection accuracy may be affected
Solution Approach 1:
The system performs preliminary action by collecting and storing comprehensive user feature data during a registration phase before privacy protection is activated. During this preliminary step, multiple features are captured and stored (facial landmarks, body characteristics), ensuring high-quality reference data is available locally on the user's device. This preliminary data collection ensures accurate matching while maintaining data security through local storage.
Solution Approach 2:
The patent applies composite materials concept by combining multiple different types of biometric features (facial features, body characteristics) into a composite feature set for each user. This multi-feature approach enhances matching accuracy by using diverse feature types that complement each other, while all features are stored securely locally on the user's device rather than relying on a single feature type.
Data Source
AI summary
Methods and systems for preserving user privacy in captured content are disclosed. A device may capture content in a location. A user located at or near the location may not want to be depicted in the content. The user may be associated with a user device comprising a privacy setting (e.g., an opted-out device). The opted-out device may receive content features associated with the content. The opted-out device may determine if the user is depicted in the content by comparing the content features to a local content feature stored on the opted-out device. If the content features correspond to the local content feature, the opted-out device may determine that the user is depicted in the content. The opted-out device may protect the privacy of the user by causing the device to obfuscate the identity of the user in the content.