Physical unclonable in-memory computing for simultaneous protecting private data and deep learning models
Abstract Compute-in-memory based on resistive random-access memory has emerged as a promising technology for accelerating neural networks on edge devices. It can reduce frequent data transfers and improve energy efficiency. However, the nonvolatile nature of resistive memory raises concerns that sto...
Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-01-01
|
| Series: | Nature Communications |
| Online Access: | https://doi.org/10.1038/s41467-025-56412-w |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832585605764087808 |
|---|---|
| author | Wenshuo Yue Kai Wu Zhiyuan Li Juchen Zhou Zeyu Wang Teng Zhang Yuxiang Yang Lintao Ye Yongqin Wu Weihai Bu Shaozhi Wang Xiaodong He Xiaobing Yan Yaoyu Tao Bonan Yan Ru Huang Yuchao Yang |
| author_facet | Wenshuo Yue Kai Wu Zhiyuan Li Juchen Zhou Zeyu Wang Teng Zhang Yuxiang Yang Lintao Ye Yongqin Wu Weihai Bu Shaozhi Wang Xiaodong He Xiaobing Yan Yaoyu Tao Bonan Yan Ru Huang Yuchao Yang |
| author_sort | Wenshuo Yue |
| collection | DOAJ |
| description | Abstract Compute-in-memory based on resistive random-access memory has emerged as a promising technology for accelerating neural networks on edge devices. It can reduce frequent data transfers and improve energy efficiency. However, the nonvolatile nature of resistive memory raises concerns that stored weights can be easily extracted during computation. To address this challenge, we propose RePACK, a threefold data protection scheme that safeguards neural network input, weight, and structural information. It utilizes a bipartite-sort coding scheme to store data with a fully on-chip physical unclonable function. Experimental results demonstrate the effectiveness of increasing enumeration complexity to 5.77 × 1075 for a 128-column compute-in-memory core. We further implement and evaluate a RePACK computing system on a 40 nm resistive memory compute-in-memory chip. This work represents a step towards developing safe, robust, and efficient edge neural network accelerators. It potentially serves as the hardware infrastructure for edge devices in federated learning or other systems. |
| format | Article |
| id | doaj-art-f119d92210fe456b82ec69ee99d8e63a |
| institution | Kabale University |
| issn | 2041-1723 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Nature Communications |
| spelling | doaj-art-f119d92210fe456b82ec69ee99d8e63a2025-01-26T12:41:17ZengNature PortfolioNature Communications2041-17232025-01-0116111310.1038/s41467-025-56412-wPhysical unclonable in-memory computing for simultaneous protecting private data and deep learning modelsWenshuo Yue0Kai Wu1Zhiyuan Li2Juchen Zhou3Zeyu Wang4Teng Zhang5Yuxiang Yang6Lintao Ye7Yongqin Wu8Weihai Bu9Shaozhi Wang10Xiaodong He11Xiaobing Yan12Yaoyu Tao13Bonan Yan14Ru Huang15Yuchao Yang16Beijing Advanced Innovation Center for Integrated Circuits, School of Integrated Circuits, Peking UniversityCollege of Electron and Information Engineering, Hebei UniversityBeijing Advanced Innovation Center for Integrated Circuits, School of Integrated Circuits, Peking UniversityBeijing Advanced Innovation Center for Integrated Circuits, School of Integrated Circuits, Peking UniversityCenter for Brain Inspired Intelligence, Chinese Institute for Brain Research (CIBR)Beijing Advanced Innovation Center for Integrated Circuits, School of Integrated Circuits, Peking UniversityBeijing Advanced Innovation Center for Integrated Circuits, School of Integrated Circuits, Peking UniversityBeijing Advanced Innovation Center for Integrated Circuits, School of Integrated Circuits, Peking UniversitySemiconductor Technology Innovation Center (Beijing) CorporationSemiconductor Technology Innovation Center (Beijing) CorporationSemiconductor Technology Innovation Center (Beijing) CorporationSemiconductor Technology Innovation Center (Beijing) CorporationCollege of Electron and Information Engineering, Hebei UniversityInstitute for Artificial Intelligence, Peking UniversityBeijing Advanced Innovation Center for Integrated Circuits, School of Integrated Circuits, Peking UniversityBeijing Advanced Innovation Center for Integrated Circuits, School of Integrated Circuits, Peking UniversityBeijing Advanced Innovation Center for Integrated Circuits, School of Integrated Circuits, Peking UniversityAbstract Compute-in-memory based on resistive random-access memory has emerged as a promising technology for accelerating neural networks on edge devices. It can reduce frequent data transfers and improve energy efficiency. However, the nonvolatile nature of resistive memory raises concerns that stored weights can be easily extracted during computation. To address this challenge, we propose RePACK, a threefold data protection scheme that safeguards neural network input, weight, and structural information. It utilizes a bipartite-sort coding scheme to store data with a fully on-chip physical unclonable function. Experimental results demonstrate the effectiveness of increasing enumeration complexity to 5.77 × 1075 for a 128-column compute-in-memory core. We further implement and evaluate a RePACK computing system on a 40 nm resistive memory compute-in-memory chip. This work represents a step towards developing safe, robust, and efficient edge neural network accelerators. It potentially serves as the hardware infrastructure for edge devices in federated learning or other systems.https://doi.org/10.1038/s41467-025-56412-w |
| spellingShingle | Wenshuo Yue Kai Wu Zhiyuan Li Juchen Zhou Zeyu Wang Teng Zhang Yuxiang Yang Lintao Ye Yongqin Wu Weihai Bu Shaozhi Wang Xiaodong He Xiaobing Yan Yaoyu Tao Bonan Yan Ru Huang Yuchao Yang Physical unclonable in-memory computing for simultaneous protecting private data and deep learning models Nature Communications |
| title | Physical unclonable in-memory computing for simultaneous protecting private data and deep learning models |
| title_full | Physical unclonable in-memory computing for simultaneous protecting private data and deep learning models |
| title_fullStr | Physical unclonable in-memory computing for simultaneous protecting private data and deep learning models |
| title_full_unstemmed | Physical unclonable in-memory computing for simultaneous protecting private data and deep learning models |
| title_short | Physical unclonable in-memory computing for simultaneous protecting private data and deep learning models |
| title_sort | physical unclonable in memory computing for simultaneous protecting private data and deep learning models |
| url | https://doi.org/10.1038/s41467-025-56412-w |
| work_keys_str_mv | AT wenshuoyue physicalunclonableinmemorycomputingforsimultaneousprotectingprivatedataanddeeplearningmodels AT kaiwu physicalunclonableinmemorycomputingforsimultaneousprotectingprivatedataanddeeplearningmodels AT zhiyuanli physicalunclonableinmemorycomputingforsimultaneousprotectingprivatedataanddeeplearningmodels AT juchenzhou physicalunclonableinmemorycomputingforsimultaneousprotectingprivatedataanddeeplearningmodels AT zeyuwang physicalunclonableinmemorycomputingforsimultaneousprotectingprivatedataanddeeplearningmodels AT tengzhang physicalunclonableinmemorycomputingforsimultaneousprotectingprivatedataanddeeplearningmodels AT yuxiangyang physicalunclonableinmemorycomputingforsimultaneousprotectingprivatedataanddeeplearningmodels AT lintaoye physicalunclonableinmemorycomputingforsimultaneousprotectingprivatedataanddeeplearningmodels AT yongqinwu physicalunclonableinmemorycomputingforsimultaneousprotectingprivatedataanddeeplearningmodels AT weihaibu physicalunclonableinmemorycomputingforsimultaneousprotectingprivatedataanddeeplearningmodels AT shaozhiwang physicalunclonableinmemorycomputingforsimultaneousprotectingprivatedataanddeeplearningmodels AT xiaodonghe physicalunclonableinmemorycomputingforsimultaneousprotectingprivatedataanddeeplearningmodels AT xiaobingyan physicalunclonableinmemorycomputingforsimultaneousprotectingprivatedataanddeeplearningmodels AT yaoyutao physicalunclonableinmemorycomputingforsimultaneousprotectingprivatedataanddeeplearningmodels AT bonanyan physicalunclonableinmemorycomputingforsimultaneousprotectingprivatedataanddeeplearningmodels AT ruhuang physicalunclonableinmemorycomputingforsimultaneousprotectingprivatedataanddeeplearningmodels AT yuchaoyang physicalunclonableinmemorycomputingforsimultaneousprotectingprivatedataanddeeplearningmodels |