What problem does this paper attempt to address?

This paper aims to address the issues of data processing efficiency and security in traditional von Neumann computing architectures, particularly in AI edge computing applications. Specifically, the paper proposes a computing-in-memory chip based on magnetic random-access memory (MRAM), which integrates complementary metal-oxide-semiconductor (CMOS) circuits to achieve efficient and secure near-memory computing. ### Main Issues: 1. **Data Processing Efficiency**: Traditional computing architectures require frequent data transfers between memory and computing units, leading to low energy efficiency and increased latency. 2. **Security**: Using AI-based edge computing devices in public areas poses security risks, such as malicious attacks and unauthorized access to sensitive data. ### Solutions: - **MRAM-based Computing-in-Memory Chip**: This chip integrates computing and storage functions, reducing the need for data transfers and improving computing efficiency. - **Physical Unclonable Function (PUF)**: The inherent randomness provided by MRAM can be used to generate highly secure PUFs, which can be used for authentication and key generation, thereby enhancing system security. ### Specific Contributions: - **High-Density and Stable Integration**: Achieved high-density and stable integration with a 6.6-megabit CMOS integrated macro manufactured using a 22-nanometer process. - **Multi-Mode Readout**: Designed a multi-mode (memory, PUF) sense amplifier that supports various data processing schemes. - **Low Power Consumption and High Reliability**: Achieved low power consumption and high reliability PUF functions through various energy efficiency optimization techniques, such as code programming, winner termination, write-back, 2D semi-complementary physical encryption, and sparse correction linear unit (ReLU) perception early termination. - **Meeting Security Requirements**: The chip exhibits uniformity, uniqueness, and reliability, making it suitable for applications requiring high reliability, such as device authentication and embedded biosensors. In summary, this paper proposes an MRAM-based computing-in-memory chip to address the challenges of data processing efficiency and security in traditional computing architectures, providing a new solution for future AI edge computing applications.

Secure computing with MRAM

Ultrafast Self-Rectifying Memristors for Advanced In-Memory Computing

A computing-in-memory macro based on three-dimensional resistive random-access memory

Phase-change Heterostructure Enables Ultralow Noise and Drift for Memory Operation.

A CMOS-integrated spintronic compute-in-memory macro for secure AI edge devices

A robust deep learning attack immune MRAM-based physical unclonable function

RRAM-Based Hardware Security Primitives

A 40nm 150 TOPS/W High Row-Parallel MRAM Compute-in-Memory Macro with Series 3T1MTJ Bitcell for MAC Operation

A 22-nm 1-Mb 1024-b Read Data-Protected STT-MRAM Macro With Near-Memory Shift-and-Rotate Functionality and 42.6-GB/s Read Bandwidth for Security-Aware Mobile Device

Concealable Physical Unclonable Function Generation and In-Memory Encryption Machine using Vertical Self-Rectifying Memristors

A Logic Resistive Memory Chip for Embedded Key Storage with Physical Security

Ultrafast Near-Ideal Phase-Change Memristive Physical Unclonable Functions Driven by Amorphous State Variations

Cnt Based Mechanical Devices for Ulsi Memory

A CMOS-integrated compute-in-memory macro based on resistive random-access memory for AI edge devices

Application of Resistive Random Access Memory in Hardware Security: A Review

A Physically-Secure Write Scheme of Multi-Time Programmable Rram for Critical Information Storage

Materials, Processes, Devices and Applications of Magnetoresistive Random Access Memory

Comprehensive Study of Security and Privacy of Emerging Non-Volatile Memories

Resistive Memory for Computing and Security: Algorithms, Architectures, and Platforms

Resistance-Sum Architecture for Voltage-Controlled SOT-MRAM based Computing-in-Memory With Hybrid References

High-density Magnetoresistive Random Access Memory Operating at Ultralow Voltage at Room Temperature