Grayscale Comparison for Lightweight Face Injection Attack Detection

Solution Overview

Problem

Current identity verification technologies, particularly in facial recognition for Internet financial scenarios, struggle to effectively detect and prevent injection attacks, which are resource-intensive and require high computing power, making them inefficient and insecure.

Innovation Solution

A lightweight injection attack detection method that utilizes grayscale value comparisons of video frames to identify differences in light changes on a target face, allowing real-time detection on mobile terminals by converting light values to grayscale and analyzing their consistency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If complex injection attack detection algorithms are used to improve detection accuracy, then security is improved, but computing resource consumption increases and response speed decreases

Engineering Contradiction:

Improveinjection attack detection accuracyVSAvoidcomputing resource consumption

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent extracts the essential detection feature from complex algorithms, isolating the grayscale value comparison as the core mechanism. By taking out only the necessary grayscale transformation and comparison operations, the system achieves effective injection attack detection while dramatically reducing computing resource consumption and enabling real-time execution on mobile terminals.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent employs simple grayscale value comparisons instead of complex, resource-intensive algorithms. This approach uses computationally inexpensive operations that can be executed rapidly and repeatedly, providing continuous real-time detection with minimal resource consumption, effectively replacing heavy computational methods with lightweight operations.

Inventive Principle:

Principle #27Cheap short-living objects (Disposable)

2Reliability

If complex injection attack detection algorithms are used to improve detection accuracy, then security is improved, but response speed decreases

Engineering Contradiction:

Improveinjection attack detection accuracyVSAvoidresponse speed

Core Design Contradiction:

ReliabilityVSSpeed

Solution Approach 1:

The patent extracts the essential detection feature from complex algorithms, isolating the grayscale value comparison as the core mechanism. By taking out only the necessary grayscale transformation and comparison operations, the system achieves effective injection attack detection while dramatically reducing computing resource consumption and enabling real-time execution on mobile terminals.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent employs simple grayscale value comparisons instead of complex, resource-intensive algorithms. This approach uses computationally inexpensive operations that can be executed rapidly and repeatedly, providing continuous real-time detection with minimal resource consumption, effectively replacing heavy computational methods with lightweight operations.

Inventive Principle:

Principle #27Cheap short-living objects (Disposable)

3Ease of operation

If grayscale value comparison is used to reduce resource consumption, then ease of operation is improved, but detection precision may worsen

Engineering Contradiction:

Improvealgorithm simplicityVSAvoiddetection precision

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The patent transforms the detection parameter from complex pixel values to grayscale values. This parameter change simplifies the computational operations while maintaining detection effectiveness. The grayscale transformation preserves the essential light change information needed to detect injection attacks, providing a balance between algorithm simplicity and detection precision.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20250322697A1Injection attack detection

Publication Date: 2025.10.16 MASHANG CONSUMER FINANCE CO LTD

AI summary

In an injection attack detection method, a video is obtained. The video includes a plurality of video frames of a target face illuminated according to a plurality of light values. A grayscale transformation is performed on a video frame of the plurality of video frames of the video to obtain a first grayscale value of the video frame. A light value of the plurality of light values is converted to obtain a second grayscale value correspond to the video frame. The video is subjected to an injection attack that is determined based on the first grayscale value and the second grayscale value.