Introduction to Confidential Proof Delegation (CPD)

We enable a prover to demonstrate possession of secret data that satisfies a public NP relation without revealing the data itself. These proofs are publicly verifiable, succinct, and computationally efficient for verifiers, making them a cornerstone of our system. However, generating proofs is resource-intensive, requiring significant computational power and memory. This poses a challenge for applications running on resource-constrained devices, such as mobile phones, and for scenarios where sensitive data must remain confidential even when proof generation is outsourced.

Fermah's Confidential Proof Delegation (CPD) protocol addresses these challenges by introducing a decentralized, privacy-preserving solution for offloading proof generation. CPD leverages a network of distributed provers and advanced techniques—specifically, over elliptic curves—to ensure that proof generation is both efficient and secure. Unlike traditional single-provers, where one entity generates the proof, CPD generalizes this concept to a multi-party setting. Here, a secret vector $\vec{w} = (w_1, \ldots, w_n)$ is distributed among N parties, and the network collaboratively produces a public proof without any single party accessing the full secret.

How CPD Enhances Privacy and Efficiency

By using CPD, users receive:

• Privacy protection while delegating proofs
• The ability to bring large computations to devices with limited processing power
• Access to Fermah's proof market for cheap and fast proof generation
• Security guarantees even if only one party is honest
• Scalable infrastructure without compromising privacy

CPD's design ensures that sensitive data remains confidential even when proof generation is delegated to Fermah's decentralized network of prover nodes. The process begins with the user's input being securely fragmented into shares, which are distributed across multiple prover nodes. These shares are constructed over elliptic curve points, enabling secure, distributed curve operations at minimal cost. The network then employs an optimized MPC protocol to collaboratively compute the proof, handling operations such as sequences of partial products and polynomial commitments with high efficiency.

For a secret input $x$, each node $i$ receives a share $s_i$ such that $\sum_{i=1}^n s_i = x \bmod p$, where $p$ is the field characteristic. The protocol ensures that no single node can reconstruct $x$ from their share alone.

The proof generation process involves computing commitments $C = g^x h^r$ where:

• $g$ and $h$ are public generators
• $x$ is the secret input
• $r$ is a random blinding factor

A key advantage of this approach is that the computational cost of proof generation becomes independent of the number of data providers. Whether one party or a hundred contribute data, the overhead remains consistent, making CPD highly scalable. Furthermore, Fermah's honest-majority MPC protocol guarantees security as long as at least one prover remains honest, and it runs in essentially the same time as a single-prover system—ensuring performance isn't sacrificed for privacy.

System Architecture

Understanding the Architecture

Fermah operates as a universal proof market with three key components:

Network Participants

Supply Side

• High-performance machines (GPUs, FPGAs)
• Initially bootstrapped by EigenLayer Operators
• Prover Nodes with GPU capabilities
• Secure computation nodes using Trusted Execution Environments (TEEs)

Demand Side

• Proof Seekers requesting generation
• Applications requiring ZK proofs
• Users with privacy-sensitive computations

Matchmaking

• Fermah Matchmaker aligns supply and demand
• Efficient resource allocation
• Dynamic pricing based on complexity

Get Started with CPD

Ready to explore CPD? Here's how to get started:

1. Contact our team to discuss your use case

Become a CPD Operator

Are you interested in running a CPD node and contributing to the network? We're looking for operators with:

• High-performance hardware (GPUs, FPGAs)
• Reliable infrastructure
• Commitment to network security
• Experience with zero-knowledge proofs