Neural Network Layer Output Data Integrity Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data processing systems for neural network computations lack effective security measures to detect tampering or corruption of output data, which can expose sensitive information such as neural network weights, especially when data is stored and processed across different layers.
Innovation Solution
A method that generates a key associated with a portion of a neural network layer, stores it, and uses it to determine if the output data has been altered by comparing it with a version of the output data intended for a subsequent neural network layer, thereby enhancing security by identifying any unauthorized modifications or corruption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If output data is stored and processed across different neural network layers, then computational efficiency is improved, but security and data integrity are worsened due to vulnerability to tampering and corruption
Solution Approach 1:
The patent applies preliminary action by generating and storing a key associated with the output data from the first neural network layer before the data is processed in subsequent layers. This key is used later to verify whether the output data has been tampered with or corrupted during storage and processing, thus preventing security issues before they can compromise the system.
2Reliability
If security measures are implemented to detect tampering of output data, then data integrity is improved, but system complexity is worsened due to additional verification mechanisms
Solution Approach 1:
The patent introduces a key as an intermediary element that mediates between the output data and the verification process. Instead of implementing complex direct verification mechanisms, the system uses this key to determine whether the output data has been tampered with, thereby maintaining data integrity while avoiding excessive system complexity.
3Productivity
If output data is transmitted between different processing units, then computational throughput is improved, but security vulnerability is worsened due to potential unauthorized access during transmission
Solution Approach 1:
The patent implements a feedback mechanism where the stored key is retrieved and used to verify the integrity of output data after it has been processed and transmitted between different neural network layers. This feedback loop allows the system to detect any unauthorized modifications that occurred during transmission or storage, thereby mitigating security vulnerabilities while maintaining computational throughput.
Data Source
AI summary
A data processing system including storage. The data processing system also includes at least one processor to generate output data using at least a portion of a first neural network layer and generate a key associated with at least the portion of the first neural network layer. The at least one processor is further operable to obtain the key from the storage and obtain a version of the output data for input into a second neural network layer. Using the key, the at least one processor is further operable to determine whether the version of the output data differs from the output data.