Daily Digest每日摘要View All查看全部

Beimel et al. characterized ideal secret sharing over small domains, proving 4-ideal structures equal F4-linear schemes and 6-ideal schemes require ideality for all domain sizes.Beimel等人在论文中解决了理想秘密共享方案在小域上的特征化问题,证明4理想结构等价于F4线性方案,6理想结构要求所有域大小均理想。
Key Points:要点:
- •Solves the long-open characterization for k=4 and k=6 ideal access structures.解决了k=4和k=6时理想访问结构的长期开放问题
- •4-ideal structures are exactly those admitting F4-linear ideal schemes.4理想结构当且仅当存在F4线性理想方案
- •6-ideal structures are exactly those ideal for all k≥2.6理想结构当且仅当对所有k≥2均理想
- •Uses new techniques linking matroids and Latin squares.利用拟阵与拉丁方的新技术建立连接
- •Shows k=6 is the most restrictive domain size.证明了6是最限制性的域大小
- •Provides new tools for constructing ideal schemes.为构造理想秘密共享方案提供了新工具

The author measures execution cost in a deterministic RISC-V interpreter and finds up to 23x wall-clock time differences for the same step count across instruction mixes and memory access patterns. Step counting alone cannot tightly bound physical cost, and performance ordering can reverse across machines, challenging protocol-level execution budgets.作者在 deterministic RISC-V 解释器中测量执行成本,发现相同步数下不同指令混合和内存访问模式可导致高达23倍的墙钟时间差异。文章强调,仅靠步数无法紧致约束物理执行成本,且跨机器性能排序会反转,对协议级执行预算提出挑战。
Key Points:要点:
- •Pointer-chasing load is 23x slower than reference load at the same step count.相同步数下,指针追逐负载的运行时间是参考负载的23倍
- •Opcode identity alone cannot distinguish sequential load from pointer-chasing, as their instruction streams are identical.仅靠指令身份无法区分顺序加载和指针追逐,因为它们指令流相同
- •Working set size affects execution time, with a cliff determined by host cache hierarchy, not intrinsic to the program.工作集大小影响执行时间,且拐点取决于宿主缓存层级而非程序本身
- •Desktop can be slower than phone at large working sets, so worst-case hardware is not necessarily low-end.桌面机在大工作集下可能比手机更慢,最坏情况硬件不一定是低端设备
- •The author invites community to run benchmarks and collect cross-CPU data to verify performance ordering reversals.作者邀请社区运行基准测试以收集跨CPU数据,验证性能排序反转
- •Scope is limited to native interpreter execution cost, not ZK proving cost.研究范围限于解释器执行成本,不涉及ZK证明成本

Zhang et al. proposed VERIF, the first dedicated ZK PIOP for IVF-Flat retrieval, verifying the integrity of RAG service retrievals.Zhang等人提出了VERIF,首个针对IVF-Flat检索的专用ZK PIOP,用于验证RAG服务中检索的完整性。
Key Points:要点:
- •VERIF is the first dedicated ZK PIOP for IVF-Flat retrievalVERIF是首个针对IVF-Flat检索的专用ZK PIOP
- •Proves top-m centroid selection, authenticated routing, exact scoring, top-k selection, and context binding证明top-m质心选择、认证路由、完整向量评分、top-k选择及上下文绑定
- •Commitment-eliding reduction keeps query-dependent scores virtual通过承诺省略归约保持查询相关分数虚拟化
- •Reduces selection claims directly to inner products over authenticated data将选择声明直接归约为认证数据上的内积
- •Unified permutation-free top-t relation with limb-decomposed range arguments handles both selection stages统一的无置换top-t关系与limb分解范围论证处理两个选择阶段
- •Prototype achieves 86.5x prover speedup and 99.1% peak memory reduction vs Plonky2原型较通用电路zkSNARK(Plonky2)实现加速86.5倍,峰值内存降低99.1%

Vitalik Buterin introduces local mixing obfuscation, which removes circuit structure through reversibility, hardening, mixing, and gadgetization.Vitalik Buterin在博客中介绍了局部混合混淆技术,通过可逆化、硬化、混合和 gadgetization 等步骤消除电路内部结构,以实现安全混淆。
Key Points:要点:
- •Local mixing obfuscation breaks circuit logic via reversibility, hardening, mixing, and gadgetization.局部混合混淆通过可逆化、硬化、混合和 gadgetization 等步骤破坏电路内部逻辑
- •Mixing includes generation mixing, splitting, and crossing walks to scramble gate/wire relations.混合步骤包括代际混合、拆分和交叉行走,旨在打乱门与线路的关系
- •Gadgetization uses nonlinear encodings to remove correlations with original circuit wires.gadgetization 使用非线性编码消除线路与原始电路的相关性
- •The construction resists linear algebra, correlation, and random bit flip attacks.方案可抵御线性代数攻击、相关攻击和随机比特翻转攻击
- •Obfuscating random circuits yields public-key encryption, extendable to general circuits.随机电路的混淆可直接构造公钥加密,并有望扩展为一般电路混淆
- •The approach is early-stage and heuristic, but introduces a novel cryptographic paradigm.作者承认方案仍处于早期阶段,依赖启发式安全,但引入了全新的密码学范式

Zhang et al. proposed MamaBearZKP, a co-design framework using 49-bit prime fields and AVX-512IFMA to boost ZKP proving throughput on modern CPUs.Zhang等人在论文中提出了MamaBearZKP框架,通过49位素数域与证明栈协同设计,利用AVX-512IFMA指令集大幅提升现代CPU上的ZKP证明吞吐量。
Key Points:要点:
- •Presents MamaBearZKP, co-designing prime fields and proving stacks using 49-bit prime MamaBear (p = 2^49 - 2^34 + 1)提出MamaBearZKP,协同设计素数域与证明栈,利用49位素数域MamaBear(p = 2^49 - 2^34 + 1)
- •Introduces systematic vectorization for AVX-512IFMA, unifying optimization of field arithmetic, protocol, and hardware针对AVX-512IFMA执行模型设计系统化向量化框架,统一优化域算术、协议与硬件原语
- •In HyperPlonk-DeepFold prover, single-thread speedups: 42x ZeroCheck, 33x ProductCheck, 15x Commit, 21x Open, 21x end-to-end在HyperPlonk-DeepFold证明器中,单线程零检查、乘积检查、承诺、打开和端到端生成分别加速42倍、33倍、15倍、21倍和21倍
- •With 8 threads, speedups reach 64x, 47x, 81x, 45x, and 45x respectively8线程执行时,对应加速最高可达64倍、47倍、81倍、45倍和45倍
- •Compared to Plonky3 with AVX-512 BabyBear backend, MamaBearZKP achieves up to 18x single-thread speedup end-to-end端到端评估中,相比已用AVX-512 BabyBear后端的Plonky3,MamaBearZKP单线程加速最高达18倍
- •Gains stem from 49-bit field headroom enabling lazy reduction, leading to fused fold-and-evaluate kernels and stay-packed dataflow性能提升源于49位域余量使能高效惰性归约,进而实现高性能融合折叠求值内核,促成统一stay-packed数据流

@MystenLabs shared how Sui chose its post-quantum signatures: ML-DSA-65 plus SLH-DSA vaults, with Falcon trade-offs, benchmarks, and a deferred zk migration.@MystenLabs 团队在博客中分享了Sui后量子签名方案的选型,选择ML-DSA-65与SLH-DSA保障高价值资产,并讨论了Falcon的取舍、性能基准和零知识迁移方案。
Key Points:要点:
- •Sui adopts ML-DSA-65 (NIST Level 3) for native signatures, adding margin against lattice attacks.采用ML-DSA-65(NIST Level 3)作为Sui原生签名,为格密码预留安全余量。
- •SLH-DSA is shipped via Move contracts, avoiding protocol upgrades and isolating lattice breaks.SLH-DSA通过Move智能合约提供,不依赖格问题,且无需协议升级即可适配外部标准。
- •Falcon was rejected for missing Level 3, unstandardized status, and hard floating-point Gaussian sampling.Falcon被否决:无Level 3、尚未标准化、浮点高斯采样实现与侧信道防护难度高。
- •Benchmarks show verification on par with Ed25519, while signature size grows ~50x; 128KB tx limit helps.基准显示验证性能与Ed25519持平,但签名体积约增50倍,交易上限128KB可容纳。
- •Private keys remain 32-byte seeds, keeping wallet backups unchanged; address aliases ease migration.私钥保持32字节种子,钱包备份不变;地址别名支持现有账户无感切换后量子密钥。
- •The team built a thin Rust wrapper over mldsa-native, favoring small auditable surface and cross-checking.团队自研Rust包装器基于Linux基金会mldsa-native,强调小型审计面与跨实现验证。

Lee et al. proposed Sluice, a read-write streaming Groth16 prover that cuts prove-phase memory from O(N) to O(log N) via the SBM algorithm. It preserves the standard interface, trading RAM for sequential storage I/O, and succeeds under an 8GB cap at N=2^23.Lee等人在论文中提出Sluice,一种读写流式Groth16证明器,通过SBM算法将证明内存从O(N)降至O(log N),在保持标准接口下以顺序存储I/O换取RAM,实验于8GB内存限制下完成N=2^23证明。
Key Points:要点:
- •Sluice reduces Groth16 prove-phase random-access memory to O(log N).Sluice将Groth16证明阶段随机访问工作内存降至O(log N)
- •Core innovation is the Split-Butterfly-Merge (SBM) streaming NTT algorithm.核心技术创新为Split-Butterfly-Merge (SBM)流式NTT算法
- •SBM uses O(log N) memory, O(N log N) total I/O, and O(log N) sequential passes.SBM仅需O(log N)内存、O(N log N)总I/O和O(log N)次顺序外存遍历
- •Combines streaming sparse R1CS evaluation with chunked Pippenger MSM.结合流式稀疏R1CS求值与分块Pippenger MSM完成证明路径
- •Fully compatible with standard Groth16: 128-byte proofs and 3-pairing verification.完全兼容标准Groth16:128字节证明、3-pairing验证,无需修改验证合约
- •Generates N=2^23 proofs under an 8GB cap, while standard provers require 16GB.在8GB内存限制下成功生成N=2^23证明,而标准证明器需16GB

Hulek et al. proposed a unified host-agent framework for benchmarking heterogeneous ZK proving systems under a common execution method, ensuring comparability via semantic contracts and validation flags, and evaluated Groth16, Plonky3, and zkVM systems.Hulek等人在论文中提出了统一的主机-代理框架,用于在共同执行方法下对异构ZK证明系统进行基准测试,通过语义合约与验证标志确保各后端在相同声明下可比,并评估了Groth16、Plonky3及zkVM等系统。
Key Points:要点:
- •The framework uses containerized agents to isolate each proving system.该框架采用容器化代理隔离运行各证明系统
- •A scheme-adapter layer interfaces with heterogeneous backends.scheme-adapter层负责适配异构证明后端
- •Semantic contracts and JSON artifacts standardize benchmark reporting.通过语义合约与JSON产物规范基准测试结果
- •Evaluated Groth16, Halo2, Plonky3, Winterfell, Nova, RISC Zero, SP1, and ICICLE.评估涵盖Groth16、Halo2、Plonky3、Winterfell、Nova、RISC Zero、SP1、ICICLE
- •On 8 CPU cores, Plonky3 and Winterfell had the best proving time and RAM.8核CPU下Plonky3与Winterfell在证明时间和内存上表现最优
- •Key methodological contribution: validate statements, disclose security profiles, and report resource limits.核心方法论贡献:验证被证陈述、披露安全配置文件、报告资源限制

Zhang et al. proposed SPEG, the first practical proof system supporting the full image transmission pipeline on personal devices. It supports JPEG compression while preserving C2PA signatures, with optimizations for non-algebraic encoding and floating-point arithmetic. Proving FHD images takes 47s (Poseidon) or 2s (fast mode). Also fixes a security gap in VIMz.张等人提出了SPEG,首个支持个人设备上完整图像传输管道的实用证明系统,支持JPEG压缩并保持C2PA签名有效性,优化了非代数JPEG编码和浮点运算。FHD图像证明耗时47秒(Poseidon模式)或2秒(快速模式),并修复了VIMz的安全漏洞。
Key Points:要点:
- •SPEG is the first practical proof system supporting the full image transmission pipeline.SPEG是首个支持完整图像传输管道的实用证明系统
- •It supports JPEG compression while verifying C2PA signatures.支持JPEG压缩同时验证C2PA签名有效性
- •Offers a Poseidon-compatible mode and a fast KZH-based mode.提供Poseidon兼容模式和基于KZH的快速模式两种方案
- •Key optimizations: moving non-algebraic JPEG encoding out of the circuit and avoiding range checks in floating-point arithmetic.关键优化包括将非代数JPEG编码移出电路及避免浮点范围检查
- •FHD proving takes 47s (Poseidon) and 2s (fast mode).FHD图像证明时间47秒(Poseidon模式)和2秒(快速模式)
- •Identifies and fixes a VIMz security gap that allows forging proofs for unauthorized images.发现并修复了VIMz中允许伪造证明的安全漏洞
Zhong et al. proposed PANDA, a scalable system using ZKPs to prove robustness and fairness of neural networks without revealing private parameters. Its core contribution is a novel algorithm for proving linear relaxation bounds for non-linear activations, enabling lightweight proofs and polynomial scaling.Zhong等人在论文中提出了PANDA系统,利用零知识证明在不泄露模型参数的前提下证明神经网络的鲁棒性与公平性,核心贡献是面向非线性激活层的新型线性松弛界证明算法,显著提升了ZKP的可扩展性。
Key Points:要点:
- •PANDA verifies model robustness and fairness with hidden parameters via ZKPs.PANDA利用ZKP在隐藏模型参数时证明模型鲁棒性和公平性。
- •Built on CROWN, a robustness certification framework.系统基于CROWN鲁棒性认证框架构建。
- •A new algorithm certifies linear relaxation bounds for non-linear layers, yielding lightweight proofs.新算法为非线性激活层证明线性松弛界限,生成轻量级证明。
- •Handles networks with 2.9M+ parameters: ~5 min to prove, ~10 s to verify.支持超过290万参数的神经网络,5分钟内生成证明,10秒验证。
- •Polynomially scalable, supporting networks 4 orders of magnitude larger than prior ZKP systems.相比先前依赖指数时间算法的系统,PANDA多项式扩展,支持网络大4个数量级。
- •Offers a new path for privacy-preserving model audits.为安全关键和合规场景中的模型审计提供新解决方案。

Samanta et al. proposed Lumora, an arithmetization-oriented family of permutation-based wide-block ciphers optimized for zkSNARK verification and post-quantum signatures.Samanta等人在论文中提出了一族面向算术化的置换型宽分组密码Lumora,用于优化zkSNARK电路中的验证效率,并支持后量子数字签名应用。
Key Points:要点:
- •Lumora adopts an AES-like SPN structure over F_{2^n} with n = 16, 32, or 64.Lumora采用AES-like SPN结构,定义在F_{2^n}上,n可取16、32、64。
- •The Even-Mansour paradigm removes a separate key schedule, cutting R1CS constraints.通过Even-Mansour范式消除独立密钥调度,减少R1CS约束数量。
- •It targets fast inside-circuit verification for zkSNARK provers.设计目标是在zkSNARK电路内实现快速验证,提升证明者性能。
- •The paper includes a detailed security analysis of the family.论文提供了详细的安全分析,涵盖差分和线性密码分析等。
- •Implementations are compared within the FAEST-EM-256 framework.实现结果通过与FAEST-EM-256框架的对比展示其效率优势。
- •The ciphers also target post-quantum signature applications.该密码族同时面向后量子签名应用,拓宽了ZK友好密码的使用场景。

Jayaraman et al. proposed TeleZK-FL, combining INT8 post-training quantization with LUT arguments to enable on-device zero-knowledge proving for verifiable federated learning, achieving ~25x faster proofs.Jayaraman等人在论文中提出了TeleZK-FL,结合INT8后训练量化与LUT参数,使边缘设备能够直接生成零知识证明,实现无需信任的联邦学习,证明速度提升约25倍。
Key Points:要点:
- •TeleZK-FL quantizes gradients to INT8, cutting communication payload by 75%.TeleZK-FL 将梯度量化为INT8,减少通信负载75%
- •Halo2 LUT arguments decouple constraint cost from arithmetic bit-width.使用Halo2 LUT参数将约束成本与位宽解耦
- •Full-model proof generation takes only 84 ms on Raspberry Pi 4.在树莓派4上生成全模型证明仅需84ms
- •AUC drops only 0.1% on CheXpert and 0.3% on PTB-XL.在CheXpert和PTB-XL上AUC仅下降0.1%和0.3%
- •Security relies on KZG commitments, classically secure but not post-quantum.安全基于KZG承诺,经典安全但非抗量子
- •Defends against large-norm gradient poisoning but not low-norm backdoors.可抵御大规模梯度中毒攻击,但对低范数后门不设防

STRK20 organized the Private Sprint hackathon, running from Aug 14 to 31, focused on building privacy apps on Starknet mainnet, with a $5,000 STRK prize pool, requiring public building and mainnet deployments.STRK20组织了Private Sprint黑客松,于8月14日至31日举行,聚焦在Starknet主网上构建隐私应用,提供5000美元STRK奖金,参赛者需公开构建并提交主网部署。
Key Points:要点:
- •Duration: Aug 14-31, 18 days in total活动时间:8月14日至31日,共18天
- •Prize: $5,000 STRK, top three get $2,500/$1,500/$1,000奖金池:$5,000 STRK,前三名分别为$2,500/$1,500/$1,000
- •Entry: via pull request adding GitHub repo and Telegram username参赛方式:通过pull request登记GitHub仓库和Telegram用户名
- •Deadline: Aug 31 23:59 UTC, winners announced Sep 4提交截止:8月31日23:59 UTC,9月4日公布结果
- •Criteria: integration depth 30%, working mainnet product 30%, innovation 25%, documentation 15%评判标准:集成深度30%,主网可用产品30%,创新25%,文档质量15%
- •Participants must deploy a real app on Starknet mainnet and build in public参赛者需在Starknet主网上部署真实应用,并公开构建

Setty instantiated Spartan over binary fields using Ligerito commitments and ring-switching, yielding a transparent, plausibly post-quantum SNARK that proves BLAKE3 at 410,000 hashes/sec on consumer hardware.Setty等人在论文中提出将Spartan实例化至二进制域,采用Ligerito承诺与ring-switching技术,实现无需可信设置的透明后量子SNARK,并在消费级硬件上达成每秒41万次BLAKE3证明。
Key Points:要点:
- •Instantiates Spartan over binary fields using the Ligerito commitment scheme将Spartan实例化至二进制域,使用Ligerito承诺方案
- •Employs Diamond and Posen's ring-switching technique采用Diamond和Posen的ring-switching技术
- •Transparent with no trusted setup; security rests solely on hash functions透明无需可信设置,安全性基于哈希函数,具备后量子潜力
- •Integrates SIMD R1CS, SuperSpartan MLE, and sum-check optimizations整合SIMD R1CS、SuperSpartan扩展及sum-check等已知优化
- •Proves BLAKE3 at 410K/sec and SHA-256 at 219K/sec on an M4 Max在M4 Max上实现410K BLAKE3与219K SHA-256每秒证明
- •Achieves 6.2ms for single SHA-256 (2KiB) in the Ethereum Foundation benchmark, the fastest in the suite以太坊基金会基准中单次SHA-256(2KiB)证明仅需6.2ms,速度最快

@drakefjustin announced EF is dropping Poseidon for SHA/BLAKE, as binary-field SNARKs like Binius and Flock make traditional hashes efficient. Targets leanVM in 2027 and deployments in 2028.@drakefjustin 宣布以太坊基金会放弃Poseidon,转向SHA或BLAKE。基于二进制域的SNARK设计(如Binius、Flock)使传统哈希在SNARK中高效可用,计划2027年推出leanVM,2028年部署。
Key Points:要点:
- •Ethereum Foundation drops Poseidon for SHA/BLAKE.以太坊基金会放弃Poseidon,转向SHA或BLAKE
- •Binius and Flock make traditional hashes SNARK-efficient via binary fields.Binius和Flock等突破性SNARK设计使传统哈希在SNARK中高效
- •Binary fields natively handle bits, reducing overhead for traditional hashes.二进制域原生处理比特,减少传统哈希的证明开销
- •1M traditional hash proofs per second on a laptop.每秒钟可在笔记本上证明100万次传统哈希调用
- •leanVM targets production in 2027, deployments in 2028.leanVM预计2027年生产就绪,2028年部署CL/DL/EL
- •Hash-based cryptography preferred over lattices/isogenies for post-quantum.哈希基密码学优于格和同源方案,增强后量子安全性

Xia et al. proposed Limber, a low-overhead integer SNARK with an integer mod-PCS, enabling generic non-native arithmetic over any PCS, achieving 67x faster RSA proofs.Xia等人在论文中提出了一种低开销整数SNARK方案Limber,通过构造整数模PCS实现任意PCS的非原生算术处理,可证明RSA算术比现有电路方案快67倍以上。
Key Points:要点:
- •Non-native arithmetic is a key SNARK bottleneck, causing high overhead and bugs.非原生算术是SNARK设计的核心瓶颈,常带来高开销和实现错误。
- •Limber is the first practical integer mod-PCS construction.论文提出Limber,首个实用的整数模PCS构造。
- •It achieves asymptotically o(1) multiplicative commitment overhead.Limber渐近上仅需o(1)乘法承诺开销。
- •It can be instantiated with any standard field PCS, including small fields.可与任何标准域多项式承诺方案实例化,包括小域。
- •A PIOP for integer R1CS over a random prime yields the full SNARK.结合随机素域上的整数R1CS PIOP构建完整SNARK。
- •RSA arithmetic is proven more than 67x faster than prior circuit-based methods.实验表明RSA算术比现有电路方法快67倍以上。

Frolov et al. propose zk-Cinema, proving video provenance in zero knowledge. They model video edits as matrix multiplications, design a SNARK-friendly video representation (sfvr), and optimize NeutronNova folding for scalable, efficient proofs.Frolov等人在论文中提出zk-Cinema,用于零知识证明视频来源。该系统将视频编辑表示为矩阵乘法,设计了SNARK友好的视频表示sfvr,并优化NeutronNova折叠方案,实现高效的大规模视频溯源证明。
Key Points:要点:
- •zk-Cinema addresses provenance for C2PA-signed edited videos.zk-Cinema针对C2PA签名视频的编辑溯源问题
- •Video edits are represented as matrix multiplications for ZK-friendly proving.将视频编辑操作转化为矩阵乘法,优化ZK证明
- •sfvr (SNARK-friendly video representation) reduces prover work.提出SNARK友好的视频表示sfvr,降低证明者工作量
- •New methods incorporate signed data into SNARK proofs.设计了将签名数据嵌入SNARK证明的新方法
- •NeutronNova is optimized for high-arity folding.优化NeutronNova折叠方案,支持高元折叠
- •Read-Write Streaming NeutronNova scales via high-performance storage and parallel computing.实现读写流式NeutronNova,利用高性能存储和并行计算扩展规模

Collomosse proposes soft redaction of image provenance via ZKPs, enabling privacy-preserving distance proofs over hidden data, such as location proximity, biometric likeness, and perceptual hash, compatible with C2PA.Collomosse在论文中提出基于零知识证明的图像来源软删除机制,可在C2PA框架下用ZKP证明隐藏属性的距离性质,兼顾隐私与信任,支持位置、生物特征和感知哈希等场景。
Key Points:要点:
- •Proposes ZKP-based soft redaction for image provenance.提出基于ZKP的图像来源软删除机制
- •Uses distance proofs for location proximity.使用距离证明支持位置邻近性验证
- •Extends to L2 distance proofs over biometric embeddings.扩展到生物特征嵌入的L2距离证明
- •Applies to perceptual hash anti-spoofing.应用于感知哈希防伪场景
- •Compatible with C2PA standard.与C2PA标准兼容
- •Proof generation in seconds, verification in milliseconds.证明生成秒级、验证毫秒级

Blockstream Research shared their parameter search for the SHRINCS post-quantum signature scheme, screening 25,935 candidates to find options ~25% smaller than SLH-DSA with lower verification cost, and open-sourced an explorer.Blockstream Research在博客中分享了为SHRINCS后量子签名方案搜索参数的过程,通过筛选25,935个候选,找到了比SLH-DSA小约25%且验证成本更低的备选方案,并发布了开源参数探索器。
Key Points:要点:
- •SHRINCS is a hash-based post-quantum signature scheme from Blockstream Research.SHRINCS是Blockstream Research提出的基于哈希的后量子签名方案
- •The search covered 25,935 parameter tuples to find practical candidates.搜索覆盖25,935个参数组合,筛选出更实用的候选
- •The best fallback signature is ~5.7KB vs 7,856 bytes for SLH-DSA.最优候选将无状态回退签名从SLH-DSA的7,856字节降至约5.7KB
- •Lower signing and verification costs compared to SLH-DSA.签名生成和验证成本均比SLH-DSA更低
- •Open-source explorer lets users set bounds and weights.开源探索器允许用户自定义约束和权重来搜索最优参数
- •Stateful options like UXMSS remain use-case dependent.状态部分(如UXMSS)按用例提供灵活性,未指定固定参数

fightfake.ai released open media-authenticity tooling that combines hardware-rooted capture signatures, edit proofs, and C2PA manifests so browsers can verify genuine capture and claimed modifications.fightfake.ai 发布开放媒体真实性工具,结合相机安全硬件签名、逐次编辑证明与C2PA清单,让浏览器验证影像是否源自真实采集,以及后续改动是否与声明一致。
Key Points:要点:
- •The embedded capture record anchors all subsequent edit proofs to the original media.采集指纹与签名嵌入影像,成为后续编辑证明所锚定的原始记录
- •fightfake-toolkit edits media, generates proofs, and packages outputs in C2PA-compatible manifests.fightfake-toolkit可编辑影像、生成对应证明,并打包为兼容C2PA的清单
- •A browser extension verifies capture signatures and checks that later edits match their claims.浏览器扩展读取凭证与证明,验证媒体在采集时已签名且后续编辑符合声明
- •C2PA preserves signed provenance, while added proofs verify pixel-level edit correctness.C2PA保留签名溯源记录,额外的编辑证明用于验证像素级改动是否正确
- •Publishers host the media, manifest, and proof file without breaking standard C2PA viewers.发布方需托管媒体、清单与证明文件;标准C2PA查看器仍保持兼容

@openvm_org released OpenVM 2.1 Preview with 64-bit RISC-V and native recompilation, enabling real-time Ethereum block proving on four RTX 5090 GPUs. It is unaudited and not recommended for production.@openvm_org 发布OpenVM 2.1预览版,引入64位RISC-V与原生重编译,使用4张RTX 5090即可实时证明以太坊主网区块;该版本尚未完成审计和形式化验证,不建议用于生产环境。
Key Points:要点:
- •Four RTX 5090 GPUs proved 7,200 mainnet blocks with 9.7s p99 and 5.0s average latency.4张RTX 5090处理7,200个主网区块,证明时间p99为9.7秒、平均5.0秒
- •Sixteen GPUs reached 3.7s p99 and 2.4s average, with live proofs submitted to EthProofs.16张GPU的p99为3.7秒、平均2.4秒,并在EthProofs实时提交证明
- •RV64IM_Zicclsm support aligns with the Ethereum Foundation zkVM target and supports Rust, C++, and Go.支持以太坊基金会zkVM目标RV64IM_Zicclsm,兼容Rust、C++与Go
- •The 64-bit word size can lower Ethereum proving cost and supports misaligned memory accesses.64位字长可降低以太坊工作负载的证明成本,并支持非对齐内存访问
- •Native recompilation emits portable C and runs 34x faster than the fast interpreter.原生重编译将guest程序转换为可移植C,执行速度比快速解释器提升34倍
- •End-to-end STARK proofs target 100-bit provable security and remain below 300kB.端到端STARK证明具备100位可证明安全性,证明大小低于300kB