Randomized Memory Refresh via Analog Noise
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing memory systems are vulnerable to algorithmic randomized attacks, where a bad actor can exploit digital components like LFSRs to induce errors across volatile memory arrays through row hammering attacks.
Innovation Solution
Implementing a memory system that performs refresh operations based on sampling a random noise signal, using an analog random noise generator to generate a random value that is mapped to a row address, thereby refreshing adjacent rows and counteracting algorithmic randomized attacks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If digital components like LFSRs are used to randomize refresh operations, then the refresh operations can be automated and controlled, but the system becomes vulnerable to algorithmic randomized attacks where bad actors can exploit the predictable patterns to induce errors
Solution Approach 1:
The patent replaces digital pseudo-random number generators (LFSRs) with an analog random noise generator. The analog generator produces truly random values through physical noise processes, which are then sampled and processed to determine refresh row addresses. This substitution eliminates the predictability of digital algorithms while maintaining automated control of refresh operations.
Solution Approach 2:
The patent changes the fundamental parameter of randomness generation from digital (algorithmic) to analog (physical noise). By sampling the analog noise signal at specific times and mapping the sampled values to row addresses, the system achieves unpredictable refresh patterns that cannot be exploited by attackers who cannot predict the physical noise values.
2Reliability
If analog random noise signals are used to determine refresh row addresses, then security against algorithmic attacks is improved, but the device complexity increases due to the need for analog noise generation and sampling circuitry
Solution Approach 1:
The patent extracts only the essential random value from the analog noise signal through sampling, rather than processing the entire continuous signal. By taking discrete samples at specific time points and mapping these sampled values to row addresses, the system achieves security with minimal analog circuitry, avoiding the need for complex analog-to-digital conversion of the full signal.
3Reliability
If refresh operations are performed based on sampled random noise values, then predictability by bad actors is reduced, but the time required to generate and process random values may increase
Solution Approach 1:
The patent performs the random value generation and sampling in advance of the actual refresh operation. The analog noise is sampled, the values are mapped to row addresses, and the refresh targets are determined before the refresh cycle begins. This preliminary action ensures that the unpredictable random values are ready when needed, avoiding time delays during critical refresh operations.
Data Source
AI summary
Methods, systems, and devices for randomized refresh operations are described. A memory system may perform refresh operations to refresh logic states stored in rows targeted by successive access operations based on sampling a random noise signal. For example, the memory system, or a host system, may implement a random noise generator configured to generate the random noise signal, and the random noise signal may be sampled to generate a random value. The random value may be mapped to a random row address, and rows adjacent to or nearby the random row may be refreshed. In some cases, the random noise generator may be implemented at the memory system, or at the host system and the random noise signal may be transmitted to the memory system. For example, the host system may perform a verification operation with the memory system prior to transmitting the random noise signal.