Yoii Inc. Releases "Trustless Werewolf" on Ethereum, Achieving Game Progression Without a "Trusted Third Party" Using Zero-Knowledge Proofs × Multi-Party Computation Technology

Yoii Inc. (Headquarters: Shibuya-ku, Tokyo; CEO: Masaharu Uno; hereinafter "Yoii") is pleased to announce the development and release of "Trustless Werewolf" on the Ethereum testnet (Sepolia). This game utilizes Yoii's unique cryptographic technology, "zk-mpc," which combines Zero-Knowledge Proofs (ZKP) and Multi-Party Computation (MPC), to enable the complete progression of a Werewolf game without requiring any trusted third party (game master).

This project has received two adoptions (February 2024 and February 2025) from the Ethereum Foundation's Ecosystem Support Program (ESP) and was realized after approximately two and a half years of research and development. All source code is open source, and its application is envisioned in a wide range of fields beyond Werewolf.

■ Background of Development

Werewolf is a prime example of an "incomplete information game" where participants hold secret information (roles such as villager, fortune teller, werewolf, etc.) and perform actions like discussion, voting, fortune-telling, and attacking while keeping their roles hidden. In traditional online Werewolf, a server operator (game master) handled all functions, including role distribution, vote tabulation, victory determination, and notification of fortune-telling results. This mechanism assumed that players would disclose their secret information to this third party and "trust" how it was handled.

However, when playing games among anonymous participants on the internet, there was no way to detect fraud or collusion by the game master. "How to achieve fair game progression in an environment without a trusted third party" is a challenge common to all aspects of society, not just games, including financial transactions, voting, and identity verification.

■ Overview of zk-mpc Technology and the Mechanism of "Trustless Werewolf"

"Trustless Werewolf" replaces all game master functions by combining the following two cryptographic technologies:

Multi-Party Computation (MPC)

Each player's secret information (role, vote target, fortune-telling target, etc.) is divided into multiple shares using a technique called "secret sharing," and each share is held by a separate computation node. Unless the nodes collude, the original information is not revealed to anyone (including the server). This allows all computations involving secret information, such as role distribution, anonymous voting, fortune-telling, attacking, and victory determination, to be executed while maintaining secrecy.

For example, in anonymous voting, only the candidate with the most votes can be calculated while keeping who voted for whom secret. Unlike traditional ZKP-only implementations where the number of votes for each candidate is disclosed, MPC allows only the result to be output while keeping the vote counts themselves secret.

Zero-Knowledge Proof (ZKP)

This is a technology for "proving that a certain computation was performed correctly, without revealing any of its content." It is important to note that ZKP is not so much a technology for protecting secrets as it is a "mechanism for allowing a third party to verify whether a computation was performed correctly."

In this system, a proof is generated that allows a third party to verify that the computation performed by MPC was correct. This proof enables anyone to verify that all game processes, such as the randomness of role distribution, the legitimacy of each action (e.g., that the person attempting to fortune-tell is indeed the fortune teller), the accuracy of voting results, and the correctness of victory determination, have not been tampered with, without disclosing internal data. This proof is generated as a "Collaborative Proof" combining MPC and ZKP.

In this way, MPC plays the role of "computing while protecting secrets," and ZKP plays the role of "proving that the computation is correct," thereby establishing a mechanism where everyone can trust the correctness of the results without anyone knowing the internal information.

Role of Blockchain (Ethereum)

Furthermore, by executing commitments (registration of encrypted role information) and ZKP verification with smart contracts on Ethereum, "third-party transparency where anyone can verify and get the same result" is achieved. In the future, it will also be possible to implement reward distribution based on game results and penalties for detecting server operator fraud.

■ Differences from Traditional Online Werewolf

The biggest difference between traditional online Werewolf and "Trustless Werewolf" is that it eliminates the need for a trusted third party and that all game processes are cryptographically verifiable.

• No one, including the server, can know other players' role information.

• Fraudulent acts such as impersonating one's role or tampering with votes are mathematically impossible.

• All game processes are verifiable with ZKP, ensuring transparency.

• Results are recorded on the blockchain, providing tamper resistance.

■ Technology Stack and Development System

MPC Protocol: SPDZ (Malicious security protocol capable of detecting fraud by malicious participants)

ZKP Scheme: Groth16 (A fast verification scheme with early proven track record in the ZKP field). The zk-mpc library also supports Marlin for generic setup.

Smart Contracts: Verification contracts deployed for each computation (role distribution, voting, victory determination, etc.). The game logic contract manages commitments.

Deployment Target: Ethereum testnet (Sepolia)

Development System: Implemented from core technology to application layer by a team of 2 (1 Researcher, 1 Engineer)

Number of Players Supported: Currently up to 9 players can play simultaneously

■ Technical Challenges and Uniqueness

The biggest technical difficulty in this project was the implementation of "Collaborative Proof," which combines MPC and ZKP. While previous research only covered arithmetic operations (addition, multiplication), complex game logic like Werewolf requires more advanced computations such as bitwise operations (AND, OR) and magnitude comparisons. The development team investigated and integrated multiple fundamental research papers to implement these computations uniquely on MPC.

Furthermore, both MPC and ZKP are computationally expensive technologies on their own, and combining them further increases the computational load. Many efforts have been made to achieve practical speeds, such as optimizing finite field conversion methods and improving performance by separating proof generation and computation.

While projects like chess and poker implemented with ZKP alone exist, this project's approach of "multiple people each holding secret information, computing while keeping it hidden, and allowing a third party to verify the correctness of the results" is highly unique on a global scale.

■ Anticipated Application Areas

"Trustless Werewolf" is a reference implementation to demonstrate the practicality of zk-mpc technology. This technology solves the general problem of "correctly computing while preserving secret information without a trusted third party, and proving its legitimacy," and is expected to be applied in a wide range of areas such as:

1. Confidential Voting / Governance

Cryptographically guarantees the accuracy of results while keeping vote content confidential. The scope of vote count disclosure can also be flexibly designed, for example, enabling advanced confidential voting where "only the candidate with the most votes is disclosed, and other vote counts remain private." This is useful for decision-making that requires consideration for future relationships, such as dismissal resolutions at corporate board meetings.

2. Confidential Auctions

Determines the highest bidder while keeping bid amounts confidential. Prevents price manipulation by organizers and fraudulent leakage of bidding information, promoting fair price formation. MPC has been used in national-scale auctions in the past.

3. Financial Transactions (Dark Pools)

Matches buyers and sellers while keeping large investors' order information confidential. Traditionally, a "trusted third party" like a securities company was required, but zk-mpc may cryptographically replace that trust.

4. Credit Scoring / Loan Screening

Integrates and computes financial data and transaction data held by multiple financial institutions without disclosing them to each other, to generate a credit score. Since the correctness of this score can be proven to a third party with ZKP, it enables both improved reliability of screening results and data privacy.

5. Identity Verification / KYC

Proves "being KYC-verified" or "being included in a specific whitelist" without disclosing personal information. The ability to integrally cross-reference authentication lists held by multiple institutions without sharing the lists themselves is a unique strength of zk-mpc that ZKP alone cannot achieve.

6. Confidential AI / Privacy-Preserving LLM

Applications to "Confidential LLM" which performs inference while keeping input data to Large Language Models (LLMs) confidential, and "ZKML" which proves that inference results were generated by an authenticated model, are anticipated. This particularly has the potential to lower barriers to AI utilization in financial and medical institutions that handle highly sensitive data.

■ Yoii's Core Business Connection and Future Outlook

Yoii is a Fintech company that provides "Yoii Fuel," a Revenue-Based Financing (RBF) option for startups and SMEs. In RBF screening, Yoii receives companies' sales data and financial information to calculate future sales forecasts and default probabilities (PD). Applying zk-mpc to these screening processes could enable scoring while keeping companies' confidential data private, and presenting the legitimacy of those results in a verifiable manner to third parties.

If such a mechanism is established, companies would be able to access funding without disclosing sensitive data, lowering the barrier to data provision. At the same time, for Yoii, it creates room to balance screening accuracy and transparency while mitigating data management risks.

Furthermore, by making the legitimacy of screening results verifiable by third parties, financial institutions and investors could also treat these judgments as a trusted common basis. This could lead to credit evaluation and risk sharing that is not dependent on specific operators, contributing to a more open financial system.

Moreover, this technology is not limited to finance but can be applied to all decision-making involving confidential data. It could also function as a platform where multiple companies or institutions can collaboratively analyze and make decisions without disclosing data.

Yoii aims to accumulate technical and operational expertise in this area, while also considering more general applications.

■ Development Team Comments

Yusuke Nakae (Researcher):

"Werewolf, as a game where multiple people hold secret information and the game progresses while keeping it hidden, was an ideal subject that required both MPC and ZKP. Unlike games like poker or chess that can be implemented with ZKP alone, Werewolf is a very special game where secret information needs to remain undisclosed until the end of the game. By developing this technology as a general-purpose library, we have opened up applications in a wide range of areas, including finance and social infrastructure."

Ryo Omori (Co-founder CTO):

"Yoii's mission is to 'increase asset liquidity and support the growth of challenging entrepreneurs and companies,' with a vision for an open financial system. We intend to incorporate the cryptographic technology cultivated through this research and development into our RBF screening for data anonymization, and eventually into blockchain-based financial infrastructure."

■ External Presentation Achievements

The research results of this project were also presented at Devcon SEA, an international Ethereum conference held in Thailand in November 2024.

■ Public Information

• zk-mpc Library (GitHub): https://github.com/Yoii-Inc/zk-mpc

• Trustless Werewolf (GitHub): https://github.com/Yoii-Inc/zk-werewolf

• Demo App: https://app.trustless-werewolf.yoii.jp

■ About Yoii Inc.

Yoii is a Fintech company that provides "Yoii Fuel," a new financing option called "Revenue-Based Financing (RBF)" for startups and SMEs. Through the provision of RBF, Yoii aims to achieve its mission of "increasing asset liquidity and supporting the growth of challenging entrepreneurs and companies."