Neural Network Hiding Apparatus Without Dedicated Encryption Unit
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for hiding images using neural networks require a dedicated processing unit for encryption, which is not always feasible, and may not meet user demands for enhanced confidentiality.
Innovation Solution
A learning apparatus and inference system that utilize a neural network to perform a hiding process without a dedicated processing unit, by training a first inference model to perform predetermined processing on input data, obtaining output data, and using this output data to train a second inference model with a processing layer for hiding, and subsequently training a third model for decryption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a dedicated encryption processing unit is added to enhance confidentiality, then the hiding capability is improved, but the device complexity increases
Solution Approach 1:
The neural network model performs multiple functions: it processes input data to generate output data, and simultaneously serves as an encryption unit by hiding the output data within the input data. This eliminates the need for a separate dedicated encryption processing unit, resolving the contradiction between enhanced confidentiality and reduced device complexity.
Solution Approach 2:
The invention merges the data processing function and the encryption function into a single neural network model. The model integrates both operations, where the same network structure that processes input data also performs the hiding/encryption operation, thereby reducing overall system complexity while maintaining strong encryption capability.
2Device complexity
If a simple neural network performs the same processing for all input images, then the device complexity is reduced, but the hiding capability deteriorates
Solution Approach 1:
The neural network model dynamically adapts its processing based on the input data characteristics. By training the model with input data specific to the target images, the network learns to optimize its parameters for that specific data type, enabling effective hiding without requiring different hardware configurations or complex manual adjustments for each image type.
Solution Approach 2:
The invention changes the parameters (weights and biases) of the neural network through training specific to the input data. This parameter adaptation allows a single simple network structure to achieve high hiding capability for specific image types, eliminating the need for complex multi-configuration systems while maintaining specialized performance.
3Reliability
If multiple hiding models are prepared for different image types, then the hiding capability is improved, but the device complexity increases
Solution Approach 1:
The neural network model is pre-trained offline with input data corresponding to the specific image type before deployment. This preliminary training action enables the single model to be optimized for specific image characteristics, eliminating the need to maintain multiple different hiding models for different image types while preserving specialized hiding capability.
Solution Approach 2:
Instead of maintaining multiple different hiding models, the invention uses a single copied model structure that is trained once with appropriate input data. The trained model can then be deployed to handle images of that type, replacing the need for multiple specialized models and reducing device complexity while maintaining hiding effectiveness.
Data Source
AI summary
A learning apparatus which is capable of implementing a hiding process even if there is no processing unit exclusively for hiding. The learning apparatus is used for the hiding process using a neural network. Output data is obtained from a first inference model that has been trained and performs predetermined processing on input data. A second inference model that includes a processing layer for hiding, which includes at least one layer for hiding the output data is obtained. The output data from the first inference model is used as input data to train the second inference model.