concrete VS AIJack

There is concrete for homomorphic encryption, but that is not really a transport/compiler (yet).

Your team is writing and maintaining a cryptographic library in Rust. You will contribute in making it fast and easy to use. This library is indeed intended for growing with new cryptographic algorithms, new hardware implementations, etc.

We are building Concrete, a fast Rust library for homomorphic encryption. https://github.com/zama-ai/concrete

Awesome. I've been keeping an eye on Zama AI, and in particular their Concrete[0] library. Glad to see they're a member now.

[0] https://github.com/zama-ai/concrete/

AIJack is easy to use and can help you secure your AI system quickly. Check our documentation for more information and start securing your AI today with AIJack. Don't wait for a hijacker to compromise your AI - take action today and protect your system with AIJack.

I have developed one OSS tool (AIJack), and I would like to ask how I manage it and where I should focus.

I want to share my project, AIJack, a security and privacy risk simulator for machine learning. Many papers show that machine learning is vulnerable to cyber-attacks and privacy violations. For example, hackers can reconstruct private training data from the trained model. To simulate such risks, AIJack allows you to experiment with various combinations of more than 30 attack and defense mechanisms, such as Model Inversion, Poisoning Attack, Evasion Attack, Federated Learning, Split Learning, Differential Privacy, and Homomorphic Encryption.

Next, we will implement FedAVG, one of the most representative methods of Federated Learning. We use AIJack, an OSS, to simulate machine learning algorithms' security and privacy risks. AIJack supports both single-process and MPI as its backend.

I have released v0.0.1-alpha of AIJack, an OSS framework to simulate various attacks and defenses against machine learning models. I have implemented more than 30 algorithms, such as Model Inversion, Poisoning Attack, Evasion Attack, Federated Learning, Split Learning, Differential Privacy, and Homomorphic Encryption. You can easily experiment with various combinations of attack and defense techniques. We will also support not only standard single-process but also MPI-backend.

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 Federated Learning, Split Learning, Differential Privacy, Homomorphic Encryption, and other heuristic approaches. I am looking forward to your feedback!

I have developed AIJack, which allows you to assess the privacy and security risks of machine learning algorithms such as Model Inversion, Poisoning Attack and Evasion Attack. AIJack also provides various defense techniques like Federated Learning, Split Learning, Differential Privacy, Homomorphic Encryption, and other heuristic approaches. You can easily experiment with various combinations of attacks and defenses.

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