Secure Image Pipeline Watermarking for Biometric Authentication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Computer vision applications, such as biometric authentication, face challenges in balancing security and processing efficiency as trusted and untrusted computing platform components have different computational powers, requiring a method to securely capture and validate images while offloading intensive computations to untrusted units.
Innovation Solution
A secure image pipeline system that stores a shared secret at the image sensor module and trusted element, generates and applies a watermark to images, and validates them using a device-specific secret value, allowing for secure image capture and delivery by utilizing both trusted and untrusted computing components efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all image processing computations are performed on trusted units, then security is improved, but processing efficiency deteriorates due to limited computational power
Solution Approach 1:
The patent divides the image processing system into trusted and untrusted units, with each performing specific functions. The trusted unit handles security-critical tasks (watermark generation, validation) while the untrusted unit handles computationally intensive tasks (image capture, preprocessing), resolving the contradiction between security and processing efficiency through functional segmentation
Solution Approach 2:
The patent introduces watermarks as an intermediary mechanism that bridges trusted and untrusted units. The watermark contains authentication information that allows the trusted unit to verify image integrity without needing to perform all computations, enabling secure outsourcing of processing to untrusted units
2Productivity
If computationally intensive tasks are offloaded to untrusted units, then processing efficiency is improved, but security deteriorates due to potential tampering
Solution Approach 1:
The patent applies watermarks to images during capture or preprocessing in the untrusted unit before the images are transmitted or further processed. This preliminary authentication embedding ensures that any subsequent tampering can be detected, allowing secure offloading of computationally intensive tasks to untrusted units
Solution Approach 2:
The trusted unit validates watermarks and provides feedback on image authenticity. This feedback mechanism allows the system to detect and reject tampered images, enabling confident offloading of processing tasks to untrusted units while maintaining security through continuous verification
3Reliability
If watermarks are made visible for authentication, then validation reliability is improved, but image quality deteriorates due to perceptible artifacts
Solution Approach 1:
The patent embeds watermarks in specific local regions of the image (such as least significant bits of pixel values or specific frequency domains) rather than uniformly across the entire image. This localized approach maintains high validation reliability while minimizing perceptible artifacts and preserving overall image quality
Data Source
AI summary
Computer-implemented methods and systems for image security in computer vision applications are described. A shared secret is stored at an image sensor module and a trusted element. The shared secret includes a device-specific secret value and a definition of a plurality of pixel locations. An image is received at the image sensor module, which generates a watermark based on the secret value and applies the watermark to the image, based on the pixel locations, to create a watermarked image. The watermarked image is received at the trusted element, which validates the watermarked image based on the watermark.