Cache Coherence Anomaly Detection in Heterogeneous Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Heterogeneous computing systems face security vulnerabilities due to unverified cache coherence protocols between hosts and accelerators, which can compromise system availability and integrity.
Innovation Solution
A cache controller at a shared last level cache dynamically detects anomalous cache coherence behavior using pair-based coherence messages as a proxy for security attack protocols, identifying and mitigating potential threats by tracking and filtering message pairs and core activities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If cache coherence protocols are implemented between host and accelerators, then system functionality and performance are improved, but security vulnerabilities and system integrity risks worsen
Solution Approach 1:
A monitoring component is introduced as an intermediary between the cache coherence protocol and the host/accelerators. This mediator monitors coherence messages (such as invalidate, update, and allocate messages) to detect anomalous patterns that indicate security attacks, allowing the system to maintain cache coherence functionality while protecting against integrity violations through dynamic analysis of message sequences and patterns
2Reliability
If security monitoring is added to detect anomalous cache coherence behavior, then system security is improved, but device complexity worsens
Solution Approach 1:
The monitoring component utilizes existing cache coherence message traffic to perform security analysis without requiring additional dedicated monitoring infrastructure. By analyzing the same coherence messages that already flow through the system for normal operation, the security monitoring achieves self-service functionality, reducing overall system complexity while maintaining security detection capabilities
Solution Approach 2:
The monitoring component is designed to handle multiple types of cache coherence messages (invalidate, update, allocate, and other proprietary messages) and detect multiple attack patterns using a unified analysis framework. This multi-functional approach allows a single monitoring system to address various security threats without requiring separate specialized components for each message type or attack vector
Data Source
AI summary
Embodiments for mitigating security vulnerabilities in a heterogeneous computing system are provided. Anomalous cache coherence behavior may be dynamically detected between a host and one or more accelerators using a cache controller at a shared last level cache based upon a pair-based coherence messages functioning as a proxy for indicating one or more security attack protocols.