benchmarking-succinct-non-interactive-proofs-for-verifiable-blockchain-state
Benchmarking Succinct Non-Interactive Proofs for Verifiable Blockchain State

Benchmarking Succinct Non-Interactive Proofs for Verifiable Blockchain State

However, direct performance comparisons across native toolchains have historically conflated protocol design with individual implementation quality. To address these inconsistencies, researchers have conducted a rigorous benchmarking study of four prominent proof systems: Groth16 via the arkworks framework, TurboPLONK via Jellyfish, a transparent Winterfell STARK, and a specialized Bulletproofs baseline. This study evaluates these implementations across workloads covering verifiable identity, batched cross-shard state transitions, and private cross-chain settlement relations.

Utilizing a standardized, deterministic update specification, the team evaluated proof performance as the number of real updates scaled from 127 to 8,191.

At the highest tested load of 8,191 updates, the Winterfell STARK integrity profile demonstrated a substantial advantage regarding computational efficiency. The implementation also excelled in resource efficiency, requiring approximately 83 times and 118 times less peak memory than Groth16 and PLONK, respectively.

These performance gains come with a notable trade-off regarding the footprint of the resulting proofs. The Winterfell implementation produced a proof of 63,706 bytes. In comparison, the pairing-based implementations of Groth16 and PLONK generated proofs of 192 bytes and 977 bytes, respectively. Consequently, the STARK-based proof was approximately 332 times larger than that of Groth16 and 65 times larger than that of PLONK, illustrating the tension between proving speed and the storage overhead inherent in transparent proof systems.

The researchers further utilized phase instrumentation to separate online proving from diagnostic satisfaction audits. While Groth16 retained the lowest relative cost in the primary modeled Pectra BLS12-381/EIP-2537 scenario, the study also incorporated a live Base Sepolia campaign to measure actual proof-payload publication. The authors emphasized that these results are implementation-stack specific, representing the performance of the software as currently configured rather than theoretical limitations of the underlying cryptographic protocols themselves.

By implementing a unified deployment framework, the study maps these empirical measurements to throughput, privacy, setup, and ledger-bound regimes. This work effectively transitions proof-system selection from a basis of folklore judgment to a measured and regime-dependent engineering decision. For system designers, the data provides an essential application-level foundation, enabling a more informed choice of non-interactive proof architecture suited to the specific technical requirements of their blockchain applications.

Subject of Research: Performance benchmarks of non-interactive proof systems for blockchain state verification

Article Title: On the performance of succinct and transparent non-interactive proof implementations for verifiable identity and cross-domain blockchain state

Article References: Tran, T.-D., Huynh, B., & Pham, V.-H. (2027). On the performance of succinct and transparent non-interactive proof implementations for verifiable identity and cross-domain blockchain state. Cluster Computing, 30(1), Article 22. https://doi.org/10.1007/s10586-026-06721-6

Image Credits: AI Generated

DOI: 10.1007/s10586-026-06721-6

Keywords: blockchain, non-interactive proofs, zero-knowledge, Groth16, TurboPLONK, Winterfell, STARK, verifiable computation, proof latency, memory efficiency, proof size, state transition, cryptography, distributed ledger technology, scaling solutions, cryptographic protocols, system engineering, EIP-2537, Base Sepolia