Kite: How to Delegate Voting Power Privately

Kamilla Nazirkhanova, Vrushank Gunjur, X. Pilli Cruz-De Jesus, Dan BoneharXiv 2025 · DOI · arXiv

Abstract

In Decentralized Autonomous Organizations (DAOs), voting power is often delegated to representatives. However, current systems like Compound’s Governor Bravo make these delegations public, creating social friction and discouraging voters from changing delegates. This paper introduces Kite, a privacy-preserving protocol for DAO voting that allows members to delegate, revoke, and re-delegate voting power without revealing their choices to the public or the delegates. By combining zk-SNARKs with additively homomorphic encryption, Kite eliminates the social cost of switching delegates while remaining efficient enough for consumer hardware and the Ethereum blockchain.

1. Introduction

In representative and liquid democracies, transparency is typically viewed as a mechanism for accountability. While it is essential to see how representatives vote, the act of choosing a representative in the crypto world is often public, leading to "social friction." If a prominent DAO member moves their delegation, it can be perceived as a public snub, discouraging voters from changing representatives even when they disagree with their performance.

Existing private voting systems often focus on the ballot itself but leave the delegation process public. Kite fills this gap by ensuring that while a delegate’s voting record can remain public for accountability, the list of individuals who empowered them remains secret.

2. Methodology and System Design

Kite utilizes a "cryptographic sandwich" to achieve private delegation without sacrificing efficiency. The protocol is built upon several key components:

  1. Additively Homomorphic Encryption (AHE): When a voter delegates, they create an encrypted vector. For a set of $N$ delegates, the vector contains $N-1$ encryptions of zero and one encryption of the voter's token balance at the index of their chosen delegate. The smart contract can sum these encrypted values to track a delegate's total power without decrypting individual choices.
  2. Zero-Knowledge SNARKs (Noir): To prevent malicious actors from delegating negative votes or exceeding their token balance, every delegation includes a zk-SNARK. This proof verifies that the encrypted vector is well-formed and that the voter owns the tokens, verified via a Merkle proof of the DAO's token distribution.
  3. Anonymity Sets: To maintain practicality, Kite introduces anonymity sets. Instead of hiding a choice among thousands of delegates, a voter hides their choice within a smaller, randomly sampled subset (e.g., 20 delegates), significantly reducing proof generation costs.
  4. Smart Contract Integration: Kite was implemented as an extension of Governor Bravo, the industry-standard voting contract.
Table 1: Global contract parameters and notation used in the Kite protocol.

As shown in Table 1, the system maintains several global parameters, including Merkle tree roots for token counts ($R_T$) and encrypted voting power ($R_{eid}$), ensuring the integrity of the delegation state at the start of an election.

3. Evaluation and Results

Kite was evaluated using a proof-of-concept implementation consisting of over 10,000 lines of code across a React frontend, Rust cryptographic servers, Noir ZK circuits, and Solidity smart contracts. Benchmarking was performed on a consumer-grade MacBook Air (M2 chip) and an Anvil testnet.

  • Practical Proving Times: For an anonymity set of 20 delegates, generating a delegation proof takes approximately 60–70 seconds. At a lower privacy level (set size of 5), this drops to 7 seconds.
  • Constant On-Chain Cost: Regardless of the anonymity set size, the gas cost to verify the proof on Ethereum remains constant at approximately 400,000 gas.
  • Deployment Efficiency: By utilizing a "master verifier" contract, the authors reduced deployment costs by over 12.5 million gas compared to individual circuit verifiers.

4. Discussion and Limitations

While Kite provides significant privacy improvements, several challenges remain. One is the "Whale Problem," where a single delegate might accumulate massive power privately, only revealed during a "nuclear" vote. The authors suggest publishing differentially private reports of delegate power to mitigate this. Additionally, if a voter privately delegates to an inactive representative, their tokens are effectively lost unless they manually undelegate.

Future work includes achieving full "receipt-freeness" to prevent coercion and exploring intermediate privacy models where a delegate's vote is revealed only to their specific delegators.

5. Conclusion

Kite demonstrates that delegation privacy and voting privacy are distinct concerns that can be addressed separately. By allowing for secret delegation alongside public voting, Kite preserves both voter privacy and delegate accountability, offering a scalable solution for modern decentralized governance.

Summary updated

BibTeX
@misc{nazirkhanova2025kite,
  author = {Kamilla Nazirkhanova and Vrushank Gunjur and X. Pilli Cruz-De Jesus and Dan Boneh},
  title = {{Kite: How to Delegate Voting Power Privately}},
  year = {2025},
  eprint = {2501.05626},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2501.05626},
}
Figure