Neural Image Obfuscation for Private Remote Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image classification technologies in distributed computing systems face security risks due to the need to transmit unencrypted images, which can be decrypted and reproduced, posing a threat to privacy and data integrity.
Innovation Solution
Obfuscate image content using a local neural network before transmission to a remote classifier, employing deterministic scrambling or deletion to create a transformed object that maintains classification capability without revealing the original image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If images are transmitted over a network for classification, then classification capability is improved, but security risk increases because images can be decrypted and reproduced
Solution Approach 1:
The patent introduces an intermediary transformation layer between the image and the classification system. A neural network transforms the input image into a transformed object that contains classification information but cannot be reverse-engineered to reconstruct the original image. This intermediary transformation resolves the contradiction by enabling classification capability while preventing image reproduction.
Solution Approach 2:
The patent changes the fundamental parameters of image data representation by applying neural network transformations. The transformed object has different mathematical properties than the original image - it preserves classification features while destroying reversible reconstruction capability. This parameter transformation allows the system to maintain security while preserving classification functionality.
2Reliability
If encryption is used to secure image transmission, then data security is improved, but the system becomes vulnerable to encryption key compromise and requires complex key management
Solution Approach 1:
Instead of copying the original image for transmission (which would expose it to decryption risks), the system creates a transformed copy that contains classification information in an irreversibly modified form. The transformed object cannot be used to reconstruct the original image, eliminating the need for complex encryption key management while maintaining data security.
3Reliability
If the local computing system performs image processing, then data privacy is improved, but processing capability is limited by local hardware resources
Solution Approach 1:
The patent segments the processing task into two parts: (1) local processing to generate the transformed object, which preserves privacy by keeping the original image local, and (2) remote classification using powerful cloud resources. This segmentation allows the system to leverage both local privacy constraints and remote computational power without compromising either aspect.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The technology described herein obfuscates image content using a local neural network and a remote neural network. The local network runs on a local computer system and a remote classifier runs in a remote computing system. Together, the local network and the remote classifier are able to classify images, while the image never leaves the local computer system. In aspects of the technology, the local network receives a local image and creates a transformed object. The transformed object may be generated by processing the image with a local neural network to generate a multidimensional array and then randomly shuffling data locations within a multidimensional array or deleting data. The deleting can be accomplished by multiplying the multidimensional array by a mask of deterministically arranged ones and zeros. The transformed object is communicated to the remote classifier in the remote computing system for classification.