Back-End Memory Control for Persistent Row Hammer Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing row hammer detection methods are probabilistic and memory-intensive, leaving vulnerabilities for data corruption, and existing row hammer detection algorithms can be bypassed by attackers, leading to data loss between windows of detection.
Innovation Solution
Implementing a control mechanism in the memory sub-system's back-end with multiple counters to track access to rows over time, using a duration between activation and pre-charge commands to determine when to refresh rows and mitigate row hammer attacks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If probabilistic row hammer detection methods are used, then detection capability is provided, but memory intensity increases and vulnerabilities for data corruption remain
Solution Approach 1:
The patent extracts the row access tracking function from the main memory system by implementing a separate, dedicated row hammer detection mechanism that operates independently. This separation allows the detection system to function with minimal memory overhead while maintaining reliable detection capability.
Solution Approach 2:
The patent introduces an intermediary counter mechanism that tracks row activations without requiring full storage of row data. The counters serve as intermediaries between the memory access requests and the detection logic, enabling efficient monitoring with reduced memory consumption.
2Reliability
If existing row hammer detection algorithms are used, then detection is provided, but attackers can bypass them leading to data loss between detection windows
Solution Approach 1:
The patent implements preliminary action by continuously tracking row access patterns and maintaining counter states between detection windows. This continuous tracking ensures that row hammer attacks are detected regardless of when they occur, eliminating the vulnerability window present in periodic detection schemes.
Solution Approach 2:
The patent employs feedback mechanisms where the counter values are continuously updated based on row activation patterns and fed back to the detection logic. This real-time feedback enables the system to adapt to attack patterns and maintain detection accuracy, preventing attackers from bypassing the detection mechanism.
3Reliability
If multiple counters are implemented to track row access over time, then data retention between detection windows is improved, but device complexity increases
Solution Approach 1:
The patent segments the counter management into distinct functional units, with each counter dedicated to specific row tracking purposes. This segmentation simplifies the overall complexity by creating modular, independent counter units that can be managed separately rather than as a monolithic complex system.
Data Source
AI summary
A control mechanism may be implemented in a back-end of a memory sub-system to refresh rows of a memory device. Rows of the memory device can be refreshed based on a quantity of times the rows have been updated in a duration of time. Rows of the memory device can also be updated based on a duration of time between receipt of the activation command for the row and a pre-charge command for the row. Row of the memory device clan further be updated utilizing a pair of counters that implement a ping pong mechanism to retain data between different consecutive durations of time.