syumei

I have developed one OSS tool (https://github.com/Koukyosyumei/AIJack), and I would like to ask how I manage it and where I should focus.

This OSS aims to simulate the security and privacy risks of Machine Learning. This tool allows users to experiment with various combinations of attacks and defenses against ML/DL models, such as Model Inversion Attacks, Poisoning Attacks, BackDoor Attacks, Differential Privacy, Encryption, K-Anonnymization, and Federated Learning. I have implemented these algorithms with Python and C++. I use black, isort, pytest, and codacy to check the quality of codes. I also generate the documentation with sphinx, but the quality is not high. The coverage of test codes is around 70%.

Since I do not have much experience with OSS projects, I don't know the best practices for OSS management and development. For example, I guess I should do the following:

- Improving the coverage of test codes

- Making documents more informative and clearer

- Writing blog posts (in many languages)

- Implementing more algorithms

- Unifying APIs for different algorithms

- Create some good-first-issuesfor new contributors

- Gathering contributors, etc

How should I prioritize them? How do you keep motivated? Some people demand a lot without gratitude or contributions (of course, most people are really lovely!). I will be a senior in April, so I must also prepare for MS or Ph.D. applications. I also have to work for my tuition. How much time do you usually spend on OSS? Should I focus on other things like research internships?

Any suggestions are really welcome —many thanks for considering my question.

AIJack allows you to assess the privacy and security risks of machine learning algorithms such as Model Inversion, Poisoning Attack, Evasion Attack, Free Rider, and Backdoor Attack. AIJack also provides various defense techniques like Differential Privacy, Homomorphic Encryption, and other heuristic approaches. In addition, AIJack provides APIs for many distributed learning schemes like Federated Learning and Split Learning. You can integrate many attack and defense methods into such collaborative learning with a few lines. We currently implement more than 30 state-of-arts methods.

https://github.com/Koukyosyumei/AIJack

I have developed a framework named AIJack to simulate various attacks against machine learning models, mainly based on PyTorch and sklearn. Currently, I have implemented more than 20 algorithms! I am looking forward to your feedback!

code: https://github.com/Koukyosyumei/AIJack

documentation: https://koukyosyumei.github.io/AIJack/intro.html