ANN Partition Integrity Verification Using Cryptographic Checks

Solution Overview

Problem

In autonomous driving systems, there is a concern for overall system safety due to potential errors or security breaches in artificial neural networks (ANNs) that could lead to unsafe operations.

Innovation Solution

The implementation of a subsystem for ANN integrity verification, which uses cryptography to check the integrity of synaptic weight and bias information, preventing actions based on inaccurate data and issuing safety warnings when integrity failures are detected.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If cryptography is used to verify ANN data integrity, then safety and reliability are improved, but system complexity and processing time increase

Engineering Contradiction:

Improvesystem safetyVSAvoidsystem complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by pre-computing and storing cryptographic verification data (hashes or digital signatures) of the original ANN weight and bias information in a secure storage location. This verification data is prepared in advance before the autonomous driving system operates, so that during runtime, the system only needs to perform relatively simple verification operations by comparing current ANN data against the pre-stored cryptographic references, rather than performing complex cryptographic computations in real-time.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent introduces cryptographic hash functions or digital signature algorithms as intermediary mechanisms between the ANN data storage and the verification process. These cryptographic intermediaries transform the original ANN parameters into compact verification tokens that can be efficiently compared, acting as a mediator that simplifies the verification task while maintaining strong security guarantees.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Reliability

If cryptography is used to verify ANN data integrity, then reliability is improved, but processing speed decreases

Engineering Contradiction:

Improvedata integrityVSAvoidsystem performance

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The patent applies preliminary action by pre-computing and storing cryptographic verification data (hashes or digital signatures) of the original ANN weight and bias information in a secure storage location. This verification data is prepared in advance before the autonomous driving system operates, so that during runtime, the system only needs to perform relatively simple verification operations by comparing current ANN data against the pre-stored cryptographic references, rather than performing complex cryptographic computations in real-time.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent employs lightweight cryptographic verification mechanisms that use computationally inexpensive hash functions or simplified digital signature verification. These verification operations are designed to be fast and resource-efficient, consuming minimal processing power and memory, thereby having negligible impact on the overall system performance while providing robust integrity verification.

Inventive Principle:

Principle #27Cheap short-living objects (Disposable)

3Reliability

If ANN integrity verification is implemented, then safety is improved, but additional computational overhead is introduced

Engineering Contradiction:

ImprovesafetyVSAvoidcomputational overhead

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent employs lightweight cryptographic verification mechanisms that use computationally inexpensive hash functions or simplified digital signature verification. These verification operations are designed to be fast and resource-efficient, consuming minimal processing power and memory, thereby having negligible impact on the overall system performance while providing robust integrity verification.

Inventive Principle:

Principle #27Cheap short-living objects (Disposable)

Data Source

PatentUS12306629B2Artificial neural network integrity verification

Publication Date: 2025.05.20 LODESTAR LICENSING GROUP LLC

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.