Tshuva et al. proposed collaborative incrementally verifiable computation in their paper, enabling multiple parties to jointly update succinct proofs during streaming computations with reduced overhead. Tshuva等人在论文中提出协作增量可验证计算,允许多方在流式计算中联合更新简洁证明,显著降低内存和通信开销。
Notes
Existing collaborative zkSNARKs suffer from high memory/communication overhead and lack updatability.
New scheme allows multiple parties to jointly update proofs at each step of streaming computation.
Constant communication overhead per step (with broadcast) and memory scaling with single step.
Applicable to privacy-preserving healthcare data aggregation, audits, and joint ML models.
Integrates IVC, folding, and MPC techniques to improve efficiency for large-scale datasets.
Breaks scalability barriers of collaborative proofs, enabling practical deployment.
现有协作zkSNARK面临内存和通信开销大、缺乏可更新性等瓶颈
新方案支持多方在流式计算中每一步联合更新证明,避免从头重算
每步计算仅需常数通信开销和单步内存开销
适用于隐私医疗数据聚合、隐私审计和联合机器学习等应用
结合IVC、折叠和MPC技术,提升大规模数据集处理效率
突破现有协作证明的扩展性限制,支持实际应用部署
零知识证明zkDaily
Q&A Deep Dive 💬今日要点 深入解析 💬
Thu星期四
03.05
2026
What is a collaborative zkSNARK and how does it differ from a traditional single-prover zkSNARK? 什么是协作式 zkSNARK,它与传统单证明者 zkSNARK 的核心区别是什么?
A collaborative zkSNARK allows multiple mutually distrustful parties to jointly produce a proof attesting to the correctness of a computation over their private inputs without revealing them. In contrast, traditional zkSNARKs assume a single prover holding the entire witness. 协作式 zkSNARK 允许多个互不信任的参与方共同生成一个证明,用来证明某个计算在所有参与者私有输入上的正确执行,而不泄露各自输入。传统 zkSNARK 通常只有单一证明者持有全部 witness 并生成证明。
What is Incrementally Verifiable Computation (IVC) and how does it help with long-running or streaming computations? 什么是 Incrementally Verifiable Computation (IVC),它如何帮助处理长时间运行或流式计算?
IVC decomposes a long computation into steps where each step updates a succinct proof that certifies both the current step and all previous steps, avoiding the need to reprove the entire computation history. IVC 将一个长计算拆分为多个步骤,每一步都会更新一个简洁证明,证明当前步骤正确并继承之前步骤的证明。因此不需要重新证明整个历史计算。
How does the construction maintain constant communication overhead per party? 该方案如何在多方环境下保持每个参与方的通信复杂度为常数?
Under a broadcast channel assumption, each party only sends a small update related to the current computation step. Since this update size does not grow with dataset size or computation history, the per-party communication remains constant. 在广播信道模型下,每个参与方只需要发送与当前步骤相关的少量证明更新信息,这些信息不会随着数据规模或历史计算长度增长,因此通信复杂度保持为常数。