Accelerator Self-Isolation for Secure Distributed TEE Termination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In a system of connected accelerator subsystems, terminating a distributed trusted execution environment (TEE) poses security risks if not done properly, as malicious code can access confidential data from unterminated accelerators due to unencrypted traffic between them.
Innovation Solution
Each accelerator subsystem is configured to self-isolate, erase workload and result data, and then terminate its TEE, ensuring that data is wiped from memory before re-enabling communication links, thus preventing malicious access to confidential information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the TEE on one accelerator is terminated without coordinating with connected accelerators, then the accelerator can be reset for new workloads, but malicious code can access confidential data from other accelerators through unencrypted traffic
Solution Approach 1:
The system performs preliminary actions by having all accelerators self-isolate and erase their confidential data from memory before terminating their TEEs. This preliminary data erasure ensures that even if one accelerator's TEE is terminated first, malicious code cannot access confidential data from other accelerators that have already erased their data or are in the process of erasing it.
Solution Approach 2:
The system converts the potential harm of unencrypted traffic between accelerators into a benefit by implementing a coordinated termination protocol. When one accelerator detects TEE termination, it triggers a cascade effect where all connected accelerators are notified to self-isolate and erase data, turning a security vulnerability into a secure termination process.
2Object-affected harmful factors
If all accelerators self-isolate and erase data before TEE termination, then security is maintained, but the process becomes more complex and time-consuming
Solution Approach 1:
Each accelerator is equipped with self-service capabilities to detect TEE termination conditions, autonomously self-isolate by disabling communication interfaces, and automatically erase confidential data from memory without requiring manual intervention or complex external coordination. This simplifies the overall system complexity while maintaining security.
Solution Approach 2:
The system implements a feedback mechanism where accelerators monitor their TEE status and communicate termination conditions to connected accelerators. When one accelerator detects TEE termination, it sends notifications to connected accelerators, which then respond by self-isolating and erasing data, creating a coordinated response through feedback loops.
3Object-affected harmful factors
If accelerators maintain encrypted traffic and TEE status, then confidential data is protected, but accelerators cannot be reset for new workloads
Solution Approach 1:
The termination process is segmented into distinct phases: detection phase (detecting TEE termination condition), isolation phase (self-isolating and erasing data), and reset phase (re-enabling communication for new workloads). This segmentation allows the system to maintain security during the critical transition period while enabling productivity after secure termination is complete.
Solution Approach 2:
The system dynamically transitions accelerators between different operational states: active TEE state (with encrypted traffic), termination state (self-isolated with data erasure), and reset state (re-enabled for new workloads). This dynamic state management allows the system to adapt to different operational requirements while maintaining security during transitions.
Data Source
AI summary
A method for securely terminating a distributed trusted execution environment spanning a plurality of work accelerators. Each accelerator is configured to self-isolate upon determining that the distributed TEE is to be terminated across the system of accelerators. The data is also wiped from the processor memory of each accelerator, such that the data cannot be read out from the processor memory once the accelerator's links are re-enabled. The self-isolation is performed on each accelerator prior to the step of terminating the TEE on that accelerator. An accelerator only re-enables its links to other accelerators once the data is wiped from its processor memory such that the secret data is removed from the accelerator memory.