You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.

README.md 2.7 kB

5 years ago
5 years ago
5 years ago
5 years ago
5 years ago
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081
  1. # MindArmour
  2. - [What is MindArmour](#what-is-mindarmour)
  3. - [Setting up](#setting-up-mindarmour)
  4. - [Docs](#docs)
  5. - [Community](#community)
  6. - [Contributing](#contributing)
  7. - [Release Notes](#release-notes)
  8. - [License](#license)
  9. [查看中文](./README_CN.md)
  10. ## What is MindArmour
  11. A tool box for MindSpore users to enhance model security and trustworthiness and protect privacy data.
  12. MindArmour model security module is designed for adversarial examples, including four submodule: adversarial examples generation, adversarial examples detection, model defense and evaluation. The architecture is shown as follow:
  13. ![mindarmour_architecture](docs/mindarmour_architecture.png)
  14. MindArmour differential privacy module Differential-Privacy implements the differential privacy optimizer. Currently, SGD, Momentum and Adam are supported. They are differential privacy optimizers based on the Gaussian mechanism.
  15. This mechanism supports both non-adaptive and adaptive policy. Rényi differential privacy (RDP) and Zero-Concentrated differential privacy(ZDP) are provided to monitor differential privacy budgets. The architecture is shown as follow:
  16. ![dp_architecture](docs/differential_privacy_architecture_en.png)
  17. ## Setting up MindArmour
  18. ### Dependencies
  19. This library uses MindSpore to accelerate graph computations performed by many machine learning models. Therefore, installing MindSpore is a pre-requisite. All other dependencies are included in `setup.py`.
  20. ### Installation
  21. #### Installation for development
  22. 1. Download source code from Gitee.
  23. ```bash
  24. git clone https://gitee.com/mindspore/mindarmour.git
  25. ```
  26. 2. Compile and install in MindArmour directory.
  27. ```bash
  28. $ cd mindarmour
  29. $ python setup.py install
  30. ```
  31. #### `Pip` installation
  32. 1. Download whl package from [MindSpore website](https://www.mindspore.cn/versions/en), then run the following command:
  33. ```
  34. pip install mindarmour-{version}-cp37-cp37m-linux_{arch}.whl
  35. ```
  36. 2. Successfully installed, if there is no error message such as `No module named 'mindarmour'` when execute the following command:
  37. ```bash
  38. python -c 'import mindarmour'
  39. ```
  40. ## Docs
  41. Guidance on installation, tutorials, API, see our [User Documentation](https://gitee.com/mindspore/docs).
  42. ## Community
  43. - [MindSpore Slack](https://join.slack.com/t/mindspore/shared_invite/enQtOTcwMTIxMDI3NjM0LTNkMWM2MzI5NjIyZWU5ZWQ5M2EwMTQ5MWNiYzMxOGM4OWFhZjI4M2E5OGI2YTg3ODU1ODE2Njg1MThiNWI3YmQ) - Ask questions and find answers.
  44. ## Contributing
  45. Welcome contributions. See our [Contributor Wiki](https://gitee.com/mindspore/mindspore/blob/master/CONTRIBUTING.md) for more details.
  46. ## Release Notes
  47. The release notes, see our [RELEASE](RELEASE.md).
  48. ## License
  49. [Apache License 2.0](LICENSE)

MindArmour关注AI的安全和隐私问题。致力于增强模型的安全可信、保护用户的数据隐私。主要包含3个模块:对抗样本鲁棒性模块、Fuzz Testing模块、隐私保护与评估模块。 对抗样本鲁棒性模块 对抗样本鲁棒性模块用于评估模型对于对抗样本的鲁棒性,并提供模型增强方法用于增强模型抗对抗样本攻击的能力,提升模型鲁棒性。对抗样本鲁棒性模块包含了4个子模块:对抗样本的生成、对抗样本的检测、模型防御、攻防评估。