Facial Recognition Spoof Detection via Pre-Stored Embeddings

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current facial recognition systems face challenges in distinguishing between real and spoofed images, particularly in environments where computational resources are limited, making it difficult to implement effective spoof detection without relying on remote servers.

Innovation Solution

The system generates and stores both live and spoof image features using neural networks, allowing for local comparison of current images against stored embeddings, thereby enhancing computational efficiency and enabling spoof detection on devices with limited resources.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If facial recognition systems implement spoof detection, then security reliability is improved, but computational resource requirements increase

Engineering Contradiction:

Improvespoof detection capabilityVSAvoidcomputational resource consumption

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The system pre-generates and stores spoof images during the enrollment phase, creating spoof embeddings in advance. When authentication is needed, the system only needs to compare the current image embedding against the pre-stored spoof embeddings, rather than generating spoof images in real-time. This preliminary preparation significantly reduces the computational burden during actual authentication while maintaining spoof detection capability.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system creates embedding copies of both live and spoof images using neural networks. These embeddings are stored in the database and used for comparison during authentication. By working with compressed embedding representations rather than full-resolution images, the system reduces computational requirements while preserving the essential features needed for spoof detection.

Inventive Principle:

Principle #26Copying

2Productivity

If spoof detection is implemented on local devices, then system autonomy and speed are improved, but device complexity increases

Engineering Contradiction:

Improveauthentication speedVSAvoidcomputational power requirement

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The system replaces complex real-time image processing mechanics with a simpler embedding comparison approach. Neural networks convert images to compact embedding vectors, and authentication becomes a matter of comparing these vectors against pre-stored spoof embeddings. This substitution of mechanical image analysis with mathematical vector comparison reduces the computational complexity required on local devices.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

By pre-generating spoof embeddings and storing them during enrollment, the system eliminates the need for complex real-time spoof image generation during authentication. The authentication process only requires comparing the current image embedding against the pre-prepared spoof embeddings, significantly simplifying the real-time computational requirements on local devices.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS12327433B2Spoof images for user authentication

Publication Date: 2025.06.10 FORD GLOBAL TECH LLC

AI summary

A facial recognition authentication system and method include obtaining a current image in response to a request for facial recognition authentication of a user. A set of current image features are generated from the input image. Live image features and spoof image features are retrieved for the user. Then a determination is output that the input image is one of live or spoofed based on a comparison of the current image features to the live image features and the spoof image features.