Secure AI Hardware Processing Logic Units for Trusted Model Execution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current AI hardware processing systems lack mechanisms for trust and security, particularly in dynamic AI solution model deployment across various stakeholders and environments, such as cloud, edge, and autonomous vehicles, making them vulnerable to data integrity issues, unauthorized access, and real-time attacks.
Innovation Solution
The implementation of a secure AI hardware system with security processing logic units (S-PLUs) and artificial intelligence processing logic units (AI-PLUs) that provide integrity verification, identity and trust establishment, secure isolation, real-time attack detection, and built-in detection mechanisms for rogue elements, ensuring secure and trusted AI model execution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If AI hardware processing systems are deployed for dynamic AI solution models across multiple stakeholders, then productivity and adaptability are improved, but security and trust are compromised
Solution Approach 1:
The patent segments the AI hardware processing system into multiple isolated virtual lanes, each dedicated to a specific stakeholder or AI model. This segmentation enables multiple stakeholders to run AI models simultaneously while maintaining security isolation, thus improving productivity without compromising trust. Each virtual lane operates independently with its own security context, preventing cross-contamination between different AI models or stakeholders.
Solution Approach 2:
The patent introduces security processing logic units (S-PLUs) as intermediary components between the AI processing logic units and the external environment. These S-PLUs act as mediators that verify security credentials, authenticate AI models, and enforce security policies before allowing AI processing to occur. This intermediary layer maintains security and trust while enabling dynamic deployment of AI models from multiple stakeholders.
2Reliability
If security verification mechanisms are added to AI hardware, then reliability is improved, but device complexity increases
Solution Approach 1:
The patent merges security processing logic units (S-PLUs) with AI processing logic units within the same hardware architecture. Instead of adding separate, independent security verification systems, the S-PLUs are integrated alongside the AI processing units, sharing common hardware resources and control structures. This merging approach improves reliability through verification mechanisms while minimizing the increase in device complexity by reusing existing hardware infrastructure.
3Reliability
If real-time attack detection is implemented, then reliability is improved, but use of energy increases
Solution Approach 1:
The patent implements self-service security mechanisms where the AI hardware system automatically detects and responds to attacks without requiring external intervention or additional energy-intensive processing. The security processing logic units continuously monitor for attack patterns using lightweight verification algorithms that operate passively during normal AI model execution. This self-service approach improves reliability through real-time attack detection while minimizing additional energy consumption by leveraging existing processing resources.
Data Source
AI summary
Aspects of the present disclosure are presented for an AI system featuring specially designed AI hardware that incorporates security features to provide iron clad trust and security to run AI applications/solution models. Presented herein are various security features for AI processing, including: a trust and integrity verifier of data during operation of an AI solution model; identity and trust establishment between an entity and the AI solution model; secure isolation for a virtual AI multilane system; a real-time attack detection and prevention mechanism; and built in detection mechanisms related to rogue security attack elements insertion during manufacturing. Aspects also include security to implement an AI network interconnecting multiple user devices in an AI environment.