hyperledger-labs / zeto

Privacy-preserving implementations of fungible and non-fungible tokens, using UTXO as the underlying transaction model
Apache License 2.0
25 stars 16 forks source link

Zeto - UTXO based privacy-preserving token toolkit using Zero Knowledge Proofs

This project hosts the multiple patterns to implement privacy preserving tokens on EVM. The patterns all share the same basic architectural concepts:

Overview of how Zeto tokens work

The following diagram illustrates the basics of Zeto tokens.

Zeto token basics

The above diagram illustrates that the secrets are transmitted from the sender to the receiver in an off-chain secure channel. Other means of sharing the secrets are avaiable in Zeto token implementations. For instance, the Zeto_AnonEnc implementation includes encrypted secrets in the transaction input, and emits an event with the encrypted values. The encrypted values can only be decrypted by the receiver.

Zeto fungible and non-fungible token implementations

The various patterns in this project use Zero Knowledge Proofs (ZKP) to demonstrate the validity of the proposed transaction. There is no centralized party to trust as in the Notary pattern, which is not implemented in this project but discussed briefly below.

Using ZKPs as validity proofs, each participant can independently submit transactions to the smart contract directly. As long as the participant is able to produce a valid proof, the transaction will be successfully verified and allowed to go through.

This project includes multiple ZKP circuits to support various privacy levels with Zeto, as listed below.

Performing key pair operations, such as deriving the public key from the private key, in the ZKP circuit requires using ZKP-friendly curves, for which we picked Babyjubjub instead of the regular Ethereum curve (secp256k1).

Another implication to the usage of ZKPs as transaction validity proof and the usage of the Babyjubjub curve, is that the signer of the transaction, eg. msg.sender, no longer bears the same significance as in other token implementations such as ERC20, ERC721, where the signer's EVM account address holds the actual assets. In Zeto tokens, it's the Babyjubjub public keys that hold the entitlement to spend the tokens. In fact, the applications are encouraged to use a different signing key for each transaction, to avoid leaking transaction behaviors and breaking anonymity.

Zeto_Anon

This is the simplest version of the ZKP circuit. Because the secrets required to open the commitment hashes, namely the output UTXO value and salt, are NOT encrypted and published as part of the transaction payload, using this version requires the secrets to be transmitted from the sender to the receiver in off-chain channels.

The statements in the proof include:

There is no history masking, meaning the associations between the consumed input UTXOs and the output UTXOs are in the clear.

Zeto_AnonEnc

This verison of the ZKP circuit adds encryption that makes it possible to provide data availability onchain. The circuit uses the sender's private key and the receiver's public key to generate a shared secret with ECDH, which guarantees that the receiver will be able to decrypt the values. The encrypted values include the value and salt of the output UTXO for the receiver. With these values the receiver is guaranteed to be able to spend the UTXO sent to them.

The statements in the proof include:

There is no history masking, meaning the association between the consumed input UTXOs and the output UTXOs are in the clear.

Zeto_AnonNullifier

To mask the association between the consumed UTXOs and the output UTXOs, we hide which UTXOs are being consumed by each transaction.

To achieve this, we employ the usage of nullifiers. It's a unique hash derived from the unique commitment it consumes. For a UTXO commitment hash(value, salt, owner public key), the nullifier is calculated as hash(value, salt, owner private key). Only the owner of the commitment can generate the nullifier hash. Each transaction will record the nullifiers in the smart contract, to ensure that they don't get re-used (double spending).

In order to prove that the UTXOs to be spent actually exist, we use a merkle tree proof inside the zero knowledge proof circuit. The merkle proof is validated against a merkle tree root that is maintained by the smart contract. The smart contract keeps track of all the new UTXOs in each transaction's output commitments array, and uses a merkle tree to calculate the root hash. Then the ZKP circuit can use a root hash as public input, to prove that the input commitments (UTXOs to be spent), which are private inputs to the circuit, are included in the merkle tree represented by the root.

The end result is that, from the onchain data, no one can figure out which UTXOs have been spent, while double spending is prevented.

The statements in the proof include:

History masking with nullifiers

Zeto_AnonNullifierKyc

The concept of "KYC with privacy" is introduced in this implementation pattern.

How to enforce a policy of "all senders and receivers of a transaction must be in a KYC registry", while maintaining anomymity of the sender and the receiver? The solution is similar to how nullifiers are supported, via merkle tree proofs.

The implementation of this pattern maintains a KYC registry in the smart contract as a Sparse Merkle Tree. The registry is maintained by a designated authority, and includes the public keys of entities that have cleared the KYC process. Each transaction must demonstrate that the public keys of the sender and the receivers are included in the KYC merkle tree, by generating a merkle proof and using it as a private input to the ZKP circuit.

The statements in the proof include:

Zeto_AnonEncNullifier

This implementation adds encryption, as described in the section above for Zeto_AnonEnc, to the pattern Zeto_AnonNullifier above.

The statements in the proof include:

Zeto_AnonEncNullifierKyc

This implementation adds encryption, as described in the section above for Zeto_AnonEnc, to the pattern Zeto_AnonNullifierKyc above.

The statements in the proof include:

Zeto_AnonEncNullifierNonRepudiation

The concept of "non-repudiation" is introduced in this implementation pattern.

Since all onchain states are hashes, with ownership information for the assets hidden, it's possible that a participant can send a transaction but subsequently deny it. Because the transaction signer account no longer reflects the identity of the asset owner, as discussed above, it will be impossible to know who was the sender of a transaction from purely looking at the onchain data, which is exactly the point for Zeto's anonymity support. This gives a malicious party the ability to gain repudiation, or deny that they were responsible for a past transaction.

This implementation pattern addresses that concern by encrypting the ownership information of each UTXO involved in a transaction with an authority's registered key. Only the designated authority will be able to decrypt the ownership information. The encryption is performed inside the ZKP circuit, thus guaranteeing that they are the actual owners of the UTXOs.

The statements in the proof include:

Zeto_NfAnon

This implements a basic non-fungible token.

For non-fungible tokens, the main concern with the transaction validity check is that the output UTXO contains the same secrets (id, uri) as the input UTXO, with only the ownership updated.

The statements in the proof include:

Zeto_NfAnonNullifier

This implements a non-fungible token using nullifiers, thus hiding the spending graph.

The statements in the proof include:

Enforce token transfer policies with a Notary

This pattern relies on a central party, called "Notary", that has access to the private states of all the parties in the system. This knowledge allows the Notary to check the validity of a proposed transaction, and enforce dynamic policies that would otherwise be difficult with some of the other approaches. Every transaction must be accompanied by a "notary certificate" that approve the proposed transaction. The certificate will be verified by the smart contract before allowing the transaction to go through.

The project does not include an implementation of a notary based token transfer policy enforcement.

Sub-projects

There are 4 sub-projects. Navigate to each sub-project to run the tests and learn how to use each library: