ANN Partition Integrity Verification Using Cryptographic Codes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In autonomous driving systems, there is a concern for overall system safety due to potential errors and security breaches that can affect the integrity of artificial neural networks (ANNs), particularly in the handling of synaptic weight and bias information, which can lead to unsafe operations if not verified.
Innovation Solution
The implementation of cryptography-based methods for ANN integrity verification, where cryptographic codes are generated and compared to ensure the accuracy of synaptic weights and biases before executing critical functions, issuing safety warnings in case of integrity failures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If cryptography-based integrity verification is implemented for ANN synaptic weights and biases, then system safety and data integrity are improved, but device complexity and computational overhead increase
Solution Approach 1:
Cryptographic verification codes (hashes) are pre-computed and stored alongside the ANN synaptic weights and biases during the model training or deployment phase. This preliminary action allows the integrity of the neural network parameters to be verified without requiring complex real-time cryptographic operations during inference, thus improving reliability while minimizing the impact on device complexity and performance.
2Measurement precision
If integrity verification operations are performed on ANN partitions, then detection precision of data integrity issues is improved, but processing time and productivity are reduced
Solution Approach 1:
The artificial neural network is divided into multiple partitions, with each partition having its own cryptographic verification code. This segmentation allows the integrity verification to be performed on individual partitions rather than the entire network at once, enabling parallel processing and reducing the overall verification time while maintaining high detection precision for data integrity issues.
3Reliability
If cryptographic codes are stored for each ANN partition, then reliability and traceability are improved, but memory usage and device complexity increase
Solution Approach 1:
Cryptographic hash functions with different output lengths (e.g., SHA-256, SHA-512) can be selected based on the required security level and available memory resources. This parameter change allows systems to adjust the balance between reliability and memory usage by choosing appropriate hash function parameters, enabling deployment on devices with varying memory constraints while maintaining adequate integrity verification capabilities.
Data Source
AI summary
An example method comprises receiving a number of inputs to a system employing an artificial neural network (ANN), wherein the ANN comprises a number of ANN partitions each having respective weight matrix data and bias data corresponding thereto stored in a memory. The method includes: determining an ANN partition to which the number of inputs correspond, reading, from the memory the weight matrix data and bias data corresponding to the determined ANN partition, and a first cryptographic code corresponding to the determined ANN partition; generating, using the weight matrix data and bias data read from the memory, a second cryptographic code corresponding to the determined ANN partition; determining whether the first cryptographic code and the second cryptographic code match; and responsive to determining a mismatch between the first cryptographic code and the second cryptographic code, issuing an indication of the mismatch to a controller of the system.