AccShield

AccShield is a confidential computing design and platform to extend Trusted Exectuion Environment (TEE) for hardware accelerators, targeting to emulate future cloud. We demonstrate the feasibility of end-to-end secure and zero-trust acceleration. In our paper we focus on the machine learning accelerators such as Tensor Processing Units (TPUs), and implemented our prototype on FPGA. But our methodology is applicable to accelerators and system integration in general.

Documentation

We are currently working on the instruction of building, running and using AccShield. Stay tuned for updates as we enhance the documentation. If you encounter any issues, bugs, or have suggestions for improvements, please don’t hesitate to reach out. We will work hard to fix them.

Publication

Please refer to our DAC’23 paper for more detailed information. If you are using AccShield in your research, the following reference is provided.

• W. Ren, W. Kozlowski, S. Koteshwara, M. Ye, H. Franke and D. Chen, "AccShield: a New Trusted Execution Environment with Machine-Learning Accelerators," 2023 60th ACM/IEEE Design Automation Conference (DAC), San Francisco, CA, USA, 2023, pp. 1-6, doi: 10.1109/DAC56929.2023.10247768.

  @INPROCE EDINGS{ren2023accshield,
      author={Ren, Wei and Kozlowski, William and Koteshwara, Sandhya and Ye, Mengmei and Franke, Hubertus and Chen, Deming},
      booktitle={2023 60th ACM/IEEE Design Automation Conference (DAC)}, 
      title={{AccShield}: a New Trusted Execution Environment with Machine-Learning Accelerators}, 
      year={2023},
      volume={},
      number={},
      pages={1-6},
      doi={10.1109/DAC56929.2023.10247768}
  }