Abstract
Abstract
Zero-knowledge proofs (ZKPs) are a powerful tool for privacy-preserving computation; however, they lack a built-in mechanism to indicate the sources of their secret inputs. This gap between verifiable computation and a trusted data source is a significant obstacle to the adoption of ZKPs in real-world applications that rely on data obtained from trusted origins. Existing approaches to bridging this gap typically require computing standard cryptographic hash functions, such as SHA-256, within the ZKP circuit. However, this process is highly inefficient and creates a serious performance bottleneck for the prover.
This thesis presents the design, implementation, and evaluation of two novel methods for efficient data source authentication within the Halo2 proving system. The first proposal, “Signing Inputs”, introduces a hybrid authentication model that uses a SNARK-optimized hash function inside the circuit and takes this as the baseline for comparison. The main contribution, “Signing Commitments”, introduces a new paradigm in which an external Data Provider signs not the data itself, but the ZKP system’s internal polynomial commitment to the secret data. This architectural shift completely eliminates the need for in-circuit hashing for authentication, thereby reducing the marginal cost to near-constant and enabling major scalability improvements. Empirical evaluations show that this method is significantly superior in terms of prover time, memory usage, and data capacity.
Ultimately, this work provides a practical and high-performance framework for developing systems that require both verifiable computation and data source authentication