SIG Introduction
AI Security SIG focuses on security issues in AI model development projects from data processing, model training, and deployment inference. The details are as follows:
- AI Model's robustness and reliability: Inference robustness to adversarial samples or natural perturbation samples. Data concept drift detection and model fault injection.
- Privacy protection: Differential privacy training and federated learning.
- Model deployment security: Model encryption and model obfuscation.
SIG code repositories
- MindAarmour.
- Federated learning's server, Federated learning's client.
SIG maintainers
- Wang Ze (Huawei)
- Xiulang (Huawei)
SIG members description
The membership of the AI Security SIG includes Members, Reviewers, Committers, and Maintainers, and is recorded in the members list. The description and application conditions of each role are as follows:
Members
- Full member of the SIG, who can participate in the daily discussions and activities of the SIG.
- Application conditions: Attended at least two SIG conferences/events.
Reviewers
- Code reviewer of the SIG code repository, responsible for reviewing the PR (pull requests) of the code repository.
- Application conditions: As a Member first, and have reviewed at least two PRs of the MindSpore main warehouse or MindArmour warehouse, and the total number of comments in the PR is greater than 10.
Committers
- Committer of the SIG code repository, who has the permission to merge code repository PRs.
- Application conditions: As a reviewer, and have merged at least five PRs to SIG code repository.
Maintainers
- Responsible for SIG's technical research direction and activity organizations.
- Application conditions: As a Committer and have passed the Maintainers qualification review (the organization will be organized after someone applies for it).
Application Method
- Fork Community repository.
- Submit a PR. Add your Gitee Home link, name (real name is not required), and email address for registering Gitee to the sig_members. Add proof of application, such as reviewed PRs and merged PRs, to the PR description. After the PR is submitted, Maintainers will review the PR.
Recent Events Preview
Previous Sessions