Privacy-Preserving Machine Face Identification with Obfuscation Markers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The process of capturing and analyzing images of human faces for identification purposes poses a risk of those images falling into the wrong hands and being used for undesirable purposes, such as tracking or unauthorized publication.
Innovation Solution
A computing system generates a modified version of an image depicting a human face by inserting an obfuscation marker that prevents human identification while including predefined symbology interpretable by machines to identify the obfuscated face, thereby allowing secure face matching and action based on the identity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If raw face images are stored for identification purposes, then face matching accuracy is improved, but security risk increases due to potential misuse of stored images
Solution Approach 1:
The patent extracts only the essential identification information from face images and stores it in the form of face vectors or biometric templates, rather than storing the complete raw images. This extraction process removes unnecessary visual data that could be misused while preserving the core identification characteristics needed for accurate face matching.
Solution Approach 2:
The patent introduces face vectors and biometric templates as intermediary representations between the original face images and the identification system. These intermediaries serve as a protective layer, allowing the system to perform accurate face matching through mathematical comparisons while preventing direct access to or misuse of the original facial images.
2Object-affected harmful factors
If face images are obfuscated to prevent identification, then privacy protection is improved, but machine-based face matching capability deteriorates
Solution Approach 1:
The patent segments the face image processing into two distinct stages: first, extracting biometric features and generating face vectors that capture essential identification characteristics; second, using these vectors for machine-based matching while the original images are obfuscated or not stored. This segmentation allows privacy protection and machine matching to operate independently without interfering with each other.
Solution Approach 2:
The patent creates a digital copy of the face identification information in the form of biometric templates and face vectors, which are mathematical representations rather than visual images. These copies contain all necessary information for accurate face matching while being inherently resistant to visual inspection or misuse, thus enabling both privacy protection and functional equivalence.
Data Source
AI summary
A method and system for use of obfuscation coding. An example method includes a computing system receiving an image that depicts a human face. The method then includes the computing system generating a modified image based on the received image, with the generating of the modified image involving inserting into the received image an obfuscation marker that (i) obfuscates the human face in a manner that prevents a person observing the modified image from identifying the human face and (ii) includes predefined symbology that is interpretable by a machine to identify the obfuscated human face. Further, the method includes using the predefined symbology in the modified image as a basis to take action based on an identity of the human face.