2PM.Network
Privacy Computing AI Model Ecosystem
Created on 16th June 2024
•
2PM.Network
Privacy Computing AI Model Ecosystem
The problem 2PM.Network solves
2PM.Network provides a comprehensive privacy computing solution that addresses several critical challenges in the digital and data-driven world:
Data Silos and Privacy
By leveraging full homomorphic encryption and federated learning within its privacy computing node framework, 2PM.Network enables multiple parties to utilize private data while ensuring its confidentiality. This approach effectively mitigates the issue of data silos, as it allows for the collaborative training of AI models without exposing the underlying data.
Accessibility and Security in Data Contribution
With a full homomorphic encryption client and a dataset specification and retrieval contract developed on 0G, 2PM.Network significantly lowers the barriers for end-users to contribute data securely.
Decentralized Verification and Security
To ensure the integrity and security of its node network, 2PM.Network will adapt a restaking protocol. This allows node operators to run verification clients that validate the data, computational processes, and results generated by the network, thus safeguarding against fraud and errors.
Monetization and Copyright of AI Models
Through the use of ERC6551, 2PM.Network copyrights AI models and distributes the inference income to those contributing privacy data and computational power.
Community-Driven Network Expansion
The planned implementation of 2PM DAO aims to further expand the network through community-driven initiatives, fostering an ecosystem that continuously evolves and scales based on collective input and incentives.
Our Deployment Proof on 0G
0G Storage Node Address:http://3.87.11.89:5678
IdentityContract: 0x799682Ef3c76f31227C2Ce9b9C55F972b6318fe2
HFLContract: 0x2608282b3c870146bA10A6cCFc3d6205a9bB8c47
DataHub: 0xcFD0dCa25a6F63592F6C98910aBBa8dd206DB39C
HLRContract: 0xd657917a0Cc2E80D8a750eD7082f692Fd471FCAe
PlonkVerifier3: 0x86D29f6d088d71AA59D6D44203e05f2ab126d852
DataRegistry: 0x0046D61859Da85620A7257aF732e8330571aC731
Challenges we ran into
During the development of our project with 0G, we encountered several challenges that significantly impacted our progress and required dedicated efforts to resolve:
Unclear Documentation and Insufficient Examples
The lack of clear testing and development documentation, coupled with insufficient examples in the SDK, posed significant hurdles. Many implementations had to be reverse-engineered from the 0G testing framework or directly from the source code. To overcome this, our team invested additional time in studying the available materials in-depth and engaged with other developers in the community through discord and discussion groups to share insights and solutions.
Unstable RPC and Deployment Issues
We frequently faced instability with the RPC interfaces, which led to numerous errors during deployment and runtime. This was particularly challenging as it hindered our ability to maintain a smooth operational flow. To address this, we implemented robust error-handling procedures and developed a set of internal tools to simulate and monitor the environment, which helped in identifying and rectifying the issues more efficiently.
Heavy Development Dependencies and Long Compilation Times
The heavy dependencies required for development resulted in prolonged compilation times. The need for various components to be integrated made troubleshooting difficult, as pinpointing the exact location of errors was often cumbersome. To mitigate these issues, we refined our development process by breaking down the build into smaller, manageable components, which allowed for parallel development and easier identification of faulty segments. We also increased our reliance on continuous integration tools to automate and streamline the build and testing phases.
Tracks Applied (2)