Secure Object Execution Across Heterogeneous Processors

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

As computer systems become more heterogeneous, incorporating various types of processors such as GPUs and AI-specific processors, there is a growing need to protect the security and integrity of applications running on these systems from malware and other forms of attacks.

Innovation Solution

The implementation of secure execution support for Artificial Intelligence (AI) systems and other heterogeneous systems, which involves creating Secure Objects that can be executed across different types of processors. These Secure Objects are protected through cryptographic means, allowing them to securely call other processors for processing while maintaining integrity and confidentiality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If Secure Objects are executed on heterogeneous processors (GPUs, AI processors), then processing capability and performance are improved, but security and integrity protection become more difficult to maintain

Engineering Contradiction:
Improveprocessing capabilityVSAvoidsecurity and integrity protection
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The system segments the heterogeneous processor architecture into trusted execution environments (secure regions) and untrusted regions. Each processor type (CPU, GPU, AI processor) can contain secure enclaves that host portions of Secure Objects, with each enclave maintaining its own cryptographic protection boundaries. This allows Secure Objects to be distributed across multiple processor types while maintaining security through spatial segmentation of trusted and untrusted code regions.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces intermediary components including a secure monitor, cryptographic authentication mechanisms, and inter-processor communication protocols that act as mediators between Secure Objects and the heterogeneous processor environment. These intermediaries verify integrity, authenticate processors, and manage secure communication channels, enabling Secure Objects to safely execute across different processor types without direct exposure to untrusted software on those processors.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If cryptographic protection is applied to Secure Objects, then security and integrity are improved, but system complexity increases

Engineering Contradiction:
Improvesecurity and integrityVSAvoidsystem complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent implements universal cryptographic primitives and authentication protocols that can be applied across all processor types in the heterogeneous system. A single secure monitor architecture and set of cryptographic mechanisms serve multiple functions: protecting Secure Objects on CPUs, GPUs, and AI processors; managing inter-processor communication security; and verifying integrity across the entire heterogeneous system. This multi-functional approach reduces overall system complexity compared to implementing separate security mechanisms for each processor type.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Adaptability or versatility

If Secure Objects can call multiple processor types, then versatility and adaptability are improved, but the risk of malware attacks increases

Engineering Contradiction:
Improveprocessor compatibilityVSAvoidmalware attack risk
Core Design Contradiction:
Adaptability or versatilityVSObject-affected harmful factors

Solution Approach 1:

The system applies preliminary anti-action through cryptographic authentication and integrity verification that occurs before Secure Objects invoke operations on heterogeneous processors. Each processor must present cryptographic proof of its trusted execution environment status before being allowed to execute Secure Object code. This pre-verification mechanism prevents malware from compromising Secure Objects by establishing security boundaries in advance, allowing the system to safely leverage the versatility of multiple processor types.

Inventive Principle:
Principle #9Preliminary anti-action

Data Source

PatentUS12204665B2Secure execution support for A.I. systems (and other heterogeneous systems)
Publication Date: 2025.01.21 INTERNATIONAL BUSINESS MACHINE CORPORATION
  • US12204665B2 patent drawing
  • US12204665B2 patent drawing
  • US12204665B2 patent drawing

AI summary

A method and apparatus for providing support for Secure Objects on a data processing system including providing a Secure Object comprising code and data that is protected on the data processing system on a first processor which is a first type of processor, wherein the data processing system includes a plurality of processors of different types, responsive to a portion of the Secure Object being needed to be executed on a second processor which is a second type of processor different than the first type of processor, by the first processor calling the second processor in a special interprocessor call, returning information by the second processor to the first processor, and retrieving, by the first processor, the information from the second processor.