Access Control
Rizenet combines simple on-chain voting with cryptographic signatures to keep the training process open to new collaborators while giving current participants full control over who can take part.
Key Ideas​
Community-Managed Admission​
Before a trainer can send updates, existing participants vote on a join request recorded in a smart contract.
- 1 vote per trainer: each approved trainer receives a single vote; a proposal is accepted once it meets a configurable “yes” percentage (e.g., 60 %).
- Token-gated voting: projects can weight votes by balances of the tokenized model-ownership asset (or another designated token), allowing larger stakeholders proportional influence.
Either method keeps the decision process transparent and verifiable on-chain.
Signed Updates​
Every model update is signed with the trainer’s private key. The aggregator:
- Verifies the signature.
- Confirms the signer’s public key is on the current whitelist.
- Records the update hash and public key on-chain for future audit and compensation tracking.
How It Works​
- Join Request: a new trainer submits its public key on-chain.
- Vote & Whitelist: current participants approve or reject via the chosen voting scheme.
- Train & Sign: approved trainers run local training, sign their model deltas, and send them to the aggregator.
- Verify & Record: the aggregator checks the signature and stores the proof on-chain.
- Compensation: validated updates count toward each contributor’s on-chain share of future model compensation.
Benefits​
- Predictable governance: clear, rule-based admission that scales with community size.
- Proven authenticity: cryptographic signatures confirm every update’s origin.
- Audit-ready history: decisions and contributions live permanently on-chain.
- Aligned incentives: only trusted trainers are eligible for compensation.
With lightweight governance and built-in provenance, Rizenet makes collaboration straightforward for industries where privacy and accountability both matter.