<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Eric Vreeland on Medium]]></title>
        <description><![CDATA[Stories by Eric Vreeland on Medium]]></description>
        <link>https://medium.com/@vreeland?source=rss-b3599ef94fdc------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*-zEK8ZLx6pXm3spG3rR0Cg.jpeg</url>
            <title>Stories by Eric Vreeland on Medium</title>
            <link>https://medium.com/@vreeland?source=rss-b3599ef94fdc------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 09:55:50 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@vreeland/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Scaling Trust in the Age of Artificial Intelligence]]></title>
            <link>https://medium.com/polyhedra-network/scaling-trust-in-the-age-of-artificial-intelligence-cab076d6164e?source=rss-b3599ef94fdc------2</link>
            <guid isPermaLink="false">https://medium.com/p/cab076d6164e</guid>
            <dc:creator><![CDATA[Eric Vreeland]]></dc:creator>
            <pubDate>Mon, 25 Nov 2024 20:32:29 GMT</pubDate>
            <atom:updated>2024-11-25T20:39:45.970Z</atom:updated>
            <content:encoded><![CDATA[<p><em>First Practical Zero-Knowledge Proof System to Verify Transformer and Large Language Models</em></p><h3>Introduction</h3><p>Trust plays a crucial role in our everyday lives, shaping our interactions in ways we often take for granted. Imagine unlocking your phone with facial recognition, applying for a loan through an AI system, or receiving an AI-powered medical diagnosis. How can you be sure these AI systems are treating you fairly and protecting your privacy? As artificial intelligence increasingly shapes our daily lives, we face a critical challenge: building trust in AI systems while protecting sensitive information on all sides.</p><p>Consider a hospital using an AI system for cancer diagnosis. The hospital needs proof the AI performs consistently and accurately, but the AI provider can’t share proprietary model details. Meanwhile, patient privacy must be protected. Or picture an autonomous vehicle accident — how do we definitively determine whether the AI or human driver was at fault? Traditional approaches to AI transparency fall short in these scenarios, as they typically require exposing sensitive details that could compromise security, privacy, or intellectual property.</p><p>Zero-knowledge machine learning (zkML) offers a breakthrough solution to these real-world challenges. Several years ago, we first introduced the notion of zero-knowledge machine learning (zkML) in our seminal work “Zero Knowledge Proofs for Decision Tree Predictions and Accuracy” [1], which demonstrated how to construct zero-knowledge proofs for both predictions and accuracy of decision tree models. This innovative technology enables mathematical verification of AI systems while keeping sensitive information private. For example, a bank can prove its loan approval AI treats all applicants fairly without revealing its proprietary algorithms. Healthcare providers can verify AI diagnostic results while protecting both patient data and model details. And in the future, zkML could provide definitive proof of decision-making processes in accidents involving AI systems, clearly establishing responsibility between human and AI agents.</p><p>The applications of zkML extend across industries. Financial institutions can demonstrate regulatory compliance of AI trading systems without exposing strategic advantages. Employers can prove their AI recruitment tools are unbiased while protecting candidate data. Smart home devices can verify they protect user privacy while delivering personalized services. Government agencies can audit AI systems for fairness and safety without compromising security.</p><p>However, implementing zkML in practice faces significant challenges. Current systems require substantial computational resources, making real-time verification difficult for complex AI models. Integration with existing AI development workflows remains cumbersome, requiring specialized expertise. While research has shown zkML’s potential with various AI architectures — from decision trees to large language models — creating practical, user-friendly tools for widespread adoption remains a work in progress.</p><h3>Summary of Innovation</h3><p>The performance of our zkML technology represents a significant leap forward in the field of secure machine learning. By achieving a performance improvement of <strong>four orders of magnitude</strong> over previous approaches, we have drastically reduced the time and computational resources required for proof generation. This enhancement is exemplified by Expander’s ability to generate proofs for the <strong>Llama-3 model with 8 billion parameters</strong> in just minutes using a high-performance CPU. Such efficiency was previously unattainable and sets a new benchmark for what’s possible in zkML applications.</p><p>Moreover, this advancement marks the <strong>first time zkML has been successfully scaled to accommodate modern transformers of this size</strong>. Scaling zkML to work with large-scale, state-of-the-art models like Llama-3 not only demonstrates the robustness of our technology but also its applicability to real-world, high-performance machine learning tasks. This breakthrough enables secure and efficient deployment of complex models, ensuring that robust fraud prevention mechanisms can be implemented without sacrificing system performance or scalability.</p><p>Besides groundbreaking performance, Polyhedra has also developed zkPyTorch, a revolutionary compiler framework for zero-knowledge machine learning that seamlessly integrates with PyTorch. This groundbreaking system represents more than just a toolchain — it’s the first comprehensive compiler framework that enables direct translation of standard PyTorch models into zero-knowledge circuits. zkPyTorch dramatically simplifies zkML implementation through automated optimization and compilation, enhancing AI model security and privacy while ensuring compliance with standards. It sets a new benchmark for secure AI deployment in real-world applications and fundamentally transforms how developers can build verifiable AI systems.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OOiM-CVQup1d81TJWbQW7w.jpeg" /></figure><p>As AI systems become more prevalent and sophisticated, the ability to verify their behavior while protecting sensitive information becomes increasingly crucial. zkML provides a mathematical foundation for building trustworthy AI systems that respect privacy, ensure consistency, and enable clear accountability.</p><p>Addressing these challenges requires more than just research — it demands a team capable of turning theory into practical solutions. At Polyhedra Network, we combine expertise in zero-knowledge proofs and AI with hands-on engineering experience to bridge this gap. Our work is carried out in close collaboration with researchers from the Berkeley Center for Responsible, Decentralized Intelligence (RDI), building on a shared legacy of innovation in ZK technology and decentralized systems.</p><p>Our work, such as verifying large language models like Llama-3 8B, showcases our ability to handle complex AI architectures and advance zkML technology. With a focus on usability and scalability, we develop solutions that bridge the gap between research and real-world applications. By integrating zkML into frameworks like PyTorch, we aim to make this technology accessible and practical for a wide range of industries.</p><h3>Zero-knowledge Machine Learning in a nutshell</h3><p>zkML leverages zero-knowledge proofs to enable verification of ML computations while preserving the privacy of models, input data, or both. Through this technique, one party can prove to another that they correctly conduct inference on a given model and a given input without revealing any sensitive information — whether that’s model parameters, input data, or intermediate computations.</p><p>Since our pioneering work [1], zkML has emerged as a fundamental approach for privacy-preserving machine learning applications, particularly in scenarios where both privacy and verifiability are essential requirements. A key functionality that zkML enables is <strong>non-interactive verification of inference</strong> — a service provider can generate a zero-knowledge proof demonstrating that a given output was legitimately produced by running the specified model on the input, whether it’s a decision tree, neural network, or other ML architecture.</p><p>Beyond verifiable inference, zkML’s capabilities extend to:</p><ul><li>AI Model Audit: Verifying a model’s performance on a specified test set through zkML, ensuring accuracy and compliance without revealing sensitive model or data details. This can be particularly useful for AI safety.</li><li>Training Data Origin Verification: Tracing and proving the lineage and authenticity of training data</li><li>Authenticated Data Labeling: Verifying the integrity of data labeling processes</li><li>Training Process Verification: Ensuring model training adheres to specified protocols and requirements.</li></ul><p>This comprehensive approach to verification addresses the full lifecycle of machine learning systems, from data preparation through model deployment and inference, establishing zkML as a cornerstone technology for trustworthy AI systems.</p><h3>Challenges and solutions</h3><p>The past four years have witnessed remarkable advancements in zkML technologies since our groundbreaking work in zkDT [1]. The field has expanded significantly, with subsequent innovations including zkCNN [2, 3, 12] and zkLLM [11], which extended zero-knowledge proof systems to neural networks and large language models respectively. Our pioneering contributions didn’t stop at theoretical foundations — we developed new, advanced distributed proving systems in [4,5,6,7] that have pushed the boundaries of what’s possible in zkML.</p><p>On the practical front, we’ve seen encouraging adoption of these ideas in real-world applications. Notable implementations include zero-knowledge proofs for nanoGPT (developed by the EZKL team [8]) and GPT-2 (developed by Modulus Labs [9]). These projects demonstrate the practical viability of zkML concepts, while also highlighting key challenges in the field.</p><p>However, significant challenges remain that our ongoing work aims to address:</p><ul><li><strong>Performance Limitations</strong>: Current proving systems like Halo2, when applied to models such as nanoGPT and GPT-2, achieve only <strong>sub-hundred</strong> parameters per second even on high-performance servers. This falls significantly short of practical requirements for real-world applications, where modern ML models contain billions of parameters. This performance gap represents a critical barrier to deploying zkML solutions at scale.</li><li><strong>Developer Accessibility</strong>: The current ecosystem requires developers to write customized circuits for ML models, demanding expertise in both cryptography and specialized programming languages. This steep learning curve presents a significant barrier to widespread adoption.</li></ul><p>These challenges underscore the importance of our continued work in advancing zkML technologies, particularly in developing more efficient proving systems and creating more developer-friendly tools and frameworks.</p><h3>Our Solutions</h3><h3>Expander</h3><p>We developed Expander, the world’s fastest ZK prover, implementing the next generation GKR protocol. With Expander we achieved an unprecedented 2.16 million Poseidon hashes per second on an AMD Ryzen 7950X3D and scales up to 16M hashes per second on 256-core server hardware [13]. This represents a transformative advancement for zkML implementations.</p><p>The performance breakthroughs are remarkable: we can prove VGG-16 with 15.2 million parameters in around 2 seconds on a single core CPU. We can also prove Llama-3 with 8 Billion parameters in around 2.7 hours on a single core CPU. Amortized, we can prove 0.8 million parameters per second, that is <strong>4 orders of magnitude faster</strong> than any number previously reported as shown in the table below.</p><p>And we didn’t stop here. We recognize that 166 minutes per 65 token inference is still far from ideal for practical applications. We are actively building hardware accelerations for our scheme [10] and developing a distributed prover infrastructure where multiple nodes can collaboratively work on the same proofs [6, 7]. Through these innovations, we are moving steadily toward our goal of <strong>real-time proving</strong> — where proof generation matches inference speed, effectively eliminating latency in the user experience.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3kw27bUcp44mkTHv12aHmQ.png" /></figure><h3>zkPyTorch</h3><p>In the ZK community, we have observed that developer accessibility remains a major obstacle to widespread adoption. Developers consistently express frustration about the need to learn new languages and understand sophisticated mathematics for zkML development. To address this challenge, we’ve developed a groundbreaking solution: a specialized compiler that works directly with PyTorch, eliminating the traditional barriers to zkML development.</p><p>Our compiler bridges the gap between AI and ZK by offering <strong>native PyTorch support</strong>, allowing developers to continue writing standard PyTorch code without learning new ZK-specific programming patterns. This seamlessly integrates with existing ML workflows while providing automated circuit optimization under the hood. The compiler intelligently translates PyTorch operations to ZK circuits and implements built-in optimizations for common ML patterns, ensuring efficient memory and computation layouts.</p><p>We implemented several proof system optimizations in our compiler. These focused primarily on faster computation verification for common mathematical operations. For example, we improved the speed of verifying Fast Fourier Transform calculations by reducing the number of required steps. We also made matrix multiplication checks more efficient through improved circuit design. These changes helped make our compiler’s verification process faster while maintaining correctness.</p><p>On the AI side, we utilize PyTorch alongside the XLA compiler to perform efficient and commonly used optimizations. Additionally, by leveraging the precomputed inference results, we significantly accelerate the verification process by treating the inference result as a prefill to the model. We also rely on the latest quantization result, which can significantly reduce the number of FP32 multiplication operations.</p><p>The end-to-end pipeline handles everything from automatic circuit generation to integration with our Expander proving system, culminating in a streamlined deployment process. This comprehensive approach dramatically reduces development time from months to days for complex models, eliminates the need for ZK expertise, and ensures code maintainability through standard PyTorch practices. Perhaps most importantly, our solution is future-proof — as ML models evolve, developers can easily update their implementations without learning new ZK-specific techniques.</p><h3>Future directions</h3><p>While our current achievements with Expander and zkPyTorch represent significant breakthroughs in zkML, we see several promising directions for further optimization and enhancement:</p><ol><li><strong>Smaller Finite Field Implementation</strong> Our current implementation uses BN254 curve, but migration to Mersenne-31 fields could potentially yield up to a 10x performance boost by reducing computational overhead in finite field operations.</li><li><strong>Multi-Machine Multi-GPU Support</strong> We are developing support for distributed computing environments leveraging multiple machines and GPUs, enabling parallel proof generation and efficient workload distribution for large-scale models.</li><li><strong>Extended Output Sequence Length</strong> While currently optimized for single-token output verification, we are developing optimizations to support efficient verification of longer output sequences, reducing per-token proving overhead and enabling practical applications requiring extended output generation.</li><li><strong>ZK-Friendly Quantization Techniques</strong> We are researching quantization methods specifically optimized for zero-knowledge proofs, focusing on minimizing arithmetic complexity in finite fields while maintaining model accuracy.</li><li><strong>Better compiler optimization </strong>Our compiler pipeline can be further optimized, potentially reducing both circuit size and proving time.</li></ol><p>These improvements will work synergistically with our existing optimizations to further reduce proving time and increase throughput. Our commitment to advancing zkML technology remains steadfast, and we welcome collaboration from the community as we work towards these</p><h3>Applications</h3><p>zkML applications span both Web2 and Web3 environments, each with distinct value propositions. In Web2 settings, zkML serves as a premium <strong>verification laye</strong>r for model providers, enabling two crucial trust guarantees. First, it provides proof to users that the correct model is being applied to the correct data, acting as a safeguard against both malicious service providers and potential model misbehavior. Second, as established in our zkDT paper [1], it offers a rigorous method to verify model accuracy by enabling providers to generate proofs of correct inference across large sets of input data. Together, these capabilities create a compelling value proposition that providers can offer as a premium service.</p><p>In Web3 environments, zkML serves a dual purpose of ensuring verifiability and privacy. On the verifiability front, we’ve pioneered the concept of <strong>Opportunistic zkML Proofs</strong>, implementing a challenge-based verification system where proofs are generated only when service providers are challenged. This approach offers numerous advantages: it reduces computational overhead during normal operation, creates strong fraud deterrence through economic incentives, lowers transaction costs when unchallenged, and enables better throughput for applications where immediate verification isn’t critical. The system provides flexible security, allowing applications to balance performance and security based on their specific needs.</p><p><strong>Privacy protection </strong>is important in systems that need to handle sensitive information<strong>.</strong> zkML makes it possible to keep data, models, and computations private while still allowing others to verify results. This ensures that both confidentiality and trust can be maintained across various applications.</p><p>In both worlds, our solution is designed with<strong> regulatory compliance</strong> as a primary consideration, implementing privacy as an opt-in feature. We’ve developed regulatory-first privacy controls that can be enabled based on specific requirements, with configurable settings to meet different regional and national regulations. The platform maintains a transparent mode by default when privacy features aren’t required, while providing clear tracking of privacy settings for regulatory reporting and audits. This flexible approach supports diverse use cases across sectors: from full transparency for public sector applications to HIPAA-compliant controls for healthcare, specialized settings for financial services, and configurable options for cross-border operations requiring compliance with multiple jurisdictional requirements.</p><h3>Looking Forward</h3><p>As we continue to develop zkML techniques alongside our Expander system, we’re committed to making AI systems more transparent, verifiable, and trustworthy, while respecting regulatory requirements and privacy needs. The zkML technology opens new possibilities for:</p><ul><li>Secure AI Model Deployment</li><li>Verifiable Training Processes</li><li>Configurable Privacy Controls</li><li>Decentralized AI Ecosystems</li><li>Novel AI Applications and Services</li></ul><p>We envision our solution as a foundational platform that will enable the next generation of verifiable AI applications. The examples shared here are just the beginning of what’s possible with this technology.</p><p>Stay tuned for more updates as we build the future of verifiable AI systems.</p><p><em>References:<br></em>[1] Zhang, J., Fang, Z., Zhang, Y., &amp; Song, D. (2020). Zero Knowledge Proofs for Decision Tree Predictions and Accuracy. In Proceedings of the 2020 ACM SIGSAC Conference on Computer and Communications Security (CCS ‘20).</p><p>[2] Feng, B., Qin, L., Zhang, Z., Ding, Y., &amp; Chu, S. (2021). ZEN: An Optimizing Compiler for Verifiable, Zero-Knowledge Neural Network Inferences. Cryptology ePrint Archive, Paper 2021/087. Available at:<a href="https://eprint.iacr.org/2021/087"> https://eprint.iacr.org/2021/087</a></p><p>[3] Liu, T., Xie, X., &amp; Zhang, Y. (2021). zkCNN: Zero Knowledge Proofs for Convolutional Neural Network Predictions and Accuracy. Cryptology ePrint Archive, Paper 2021/673. Available at:<a href="https://eprint.iacr.org/2021/673"> https://eprint.iacr.org/2021/673</a></p><p>[4] Xie, T., Zhang, J., Zhang, Y., Papamanthou, C., &amp; Song, D. (2019). Libra: Succinct Zero-Knowledge Proofs with Optimal Prover Computation. Cryptology ePrint Archive, Paper 2019/317. Available at:<a href="https://eprint.iacr.org/2019/317"> https://eprint.iacr.org/2019/317</a></p><p>[5] Zhang, J., Xie, T., Zhang, Y., &amp; Song, D. (2019). Transparent Polynomial Delegation and Its Applications to Zero Knowledge Proof. Cryptology ePrint Archive, Paper 2019/1482. Available at:<a href="https://eprint.iacr.org/2019/1482"> https://eprint.iacr.org/2019/1482</a></p><p>[6] Liu, T., Xie, T., Zhang, J., Song, D., &amp; Zhang, Y. (2023). Pianist: Scalable zkRollups via Fully Distributed Zero-Knowledge Proofs. Cryptology ePrint Archive, Paper 2023/1271. Available at:<a href="https://eprint.iacr.org/2023/1271"> https://eprint.iacr.org/2023/1271</a></p><p>[7] Xie, T., Zhang, J., Cheng, Z., Zhang, F., Zhang, Y., Jia, Y., Boneh, D., &amp; Song, D. (2022). zkBridge: Trustless Cross-chain Bridges Made Practical. In Proceedings of the 2022 ACM Conference on Computer and Communications Security (CCS ‘22).</p><p>[8] Modulus Labs. (2024). Chapter 14: The World’s 1st On-Chain LLM. Medium. Available at:<a href="https://medium.com/@ModulusLabs/chapter-14-the-worlds-1st-on-chain-llm-7e389189f85e"> https://medium.com/@ModulusLabs/chapter-14-the-worlds-1st-on-chain-llm-7e389189f85e</a></p><p>[9] EZKL. (2023). Honey I SNARKED the GPT. Available at:<a href="https://medium.com/@ModulusLabs/chapter-14-the-worlds-1st-on-chain-llm-7e389189f85e"> </a><a href="https://blog.ezkl.xyz/post/nanogpt/">https://blog.ezkl.xyz/post/nanogpt/</a></p><p>[10] Polyhedra Network. “Accelerating Ethereum’s Verge: Harnessing GPU Power for Zero-Knowledge Proofs.” GitHub Repository, PolyhedraZK/blogs, 2024.<a href="https://github.com/PolyhedraZK/blogs/blob/main/blogs/sumcheck_cuda.md"> https://github.com/PolyhedraZK/blogs/blob/main/blogs/sumcheck_cuda.md</a>.</p><p>[11] Sun, H., Li, J., &amp; Zhang, H. (2023). “zkLLM: Zero Knowledge Proofs for Large Language Models.”</p><p>[12] D. Kang, T. Hashimoto, I. Stoica, and Y. Sun, “Scaling up Trustless DNN Inference with Zero-Knowledge Proofs,” arXiv preprint, Oct. 2022. [Online]. Available: arXiv:2210.08674</p><p>[13] Expander performance. <a href="https://medium.com/polyhedra-network/expander-still-the-worlds-fastest-zk-prover-6aa5fd428609">https://medium.com/polyhedra-network/expander-still-the-worlds-fastest-zk-prover-6aa5fd428609</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cab076d6164e" width="1" height="1" alt=""><hr><p><a href="https://medium.com/polyhedra-network/scaling-trust-in-the-age-of-artificial-intelligence-cab076d6164e">Scaling Trust in the Age of Artificial Intelligence</a> was originally published in <a href="https://medium.com/polyhedra-network">Polyhedra Network</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Berkeley RDI & Polyhedra Unveil Production-Ready zkML: 4 Years After Pioneering the Concept]]></title>
            <link>https://medium.com/polyhedra-network/berkeley-rdi-polyhedra-unveil-production-ready-zkml-4-years-after-pioneering-the-concept-283d03b990a5?source=rss-b3599ef94fdc------2</link>
            <guid isPermaLink="false">https://medium.com/p/283d03b990a5</guid>
            <dc:creator><![CDATA[Eric Vreeland]]></dc:creator>
            <pubDate>Tue, 19 Nov 2024 17:00:40 GMT</pubDate>
            <atom:updated>2024-11-19T17:00:40.529Z</atom:updated>
            <content:encoded><![CDATA[<p>Trustlessness has always been the focus of research and development at Polyhedra. Whether intentional or not, humans are responsible for the majority of incidents involving technology. “We founded Polyhedra to build systems that can operate without human intervention, that are verified by math, and are cryptographically secure,” CTO Tiancheng Xie shared.</p><p>Four years ago, when artificial intelligence was just starting to become mainstream, our team was quick to acknowledge the power of AI. But with great power comes great responsibility, and we also realized that without proper precautions in place the opaque nature of AI could lead to disastrous outcomes as unverifiable outputs are used to make critical decisions.</p><p>That’s why, in 2020 as part of Berkeley RDI our Chief Scientist, Jiaheng Zhang, advisors Yupeng Zhang and Dawn Song published “Zero Knowledge Proofs for Decision Tree Predictions and Accuracy.” In this paper we defined for the first time the idea of zkML and the ability to prove that a model computes its prediction on a specific data sample or achieves a certain accuracy on a dataset without revealing any information about the model itself.</p><figure><img alt="verifiable AI" src="https://cdn-images-1.medium.com/max/1011/1*0G3AXNz_XWyqGUwmkQUi2A.png" /></figure><p>At the time, our research was theoretical. Four years later, we are excited to announce that we have developed a <a href="https://medium.com/polyhedra-network/scaling-trust-in-the-age-of-artificial-intelligence-3af459deee03">production-ready compiler that lets any AI developers utilize zkML</a> without any expertise in zero-knowledge proofs.</p><h4>What is zkML?</h4><p>zkML, a subset of verifiable computation, addresses trust in the context of AI by enabling proof of the correctness of model inference. Through zero-knowledge proofs, a service provider can demonstrate that a specific output was genuinely produced by running a given model on an input. Whether it’s a decision tree, neural network, or another machine learning model, zkML ensures verifiability without exposing underlying data or models.</p><h4>Extended Capabilities of zkML</h4><p>The potential of zkML extends beyond simple inference verification. It can be applied to:</p><ul><li><strong>Data Origin Verification</strong>: Tracing and proving the lineage and authenticity of training data.</li><li><strong>Authenticated Data Labeling</strong>: Ensuring that data labeling processes are genuine and unaltered.</li><li><strong>Training Process Verification</strong>: Demonstrating that a model’s training adhered to predefined protocols and requirements.</li></ul><p>Prior work on zkML has been purely theoretical due to the high computational demands; however, advancements in zero-knowledge technology now make production-ready solutions a reality. <strong>Expander proof system</strong>, which has broken many prover performance records, is the foundation of our zkML technology. This progress paves the way for zkML applications that can build trust without compromising on privacy or compliance.</p><p>“Berkeley RDI and Polyhedra are setting a new standard for trust and transparency in artificial intelligence with innovative zero-knowledge machine learning (zkML) technology, a groundbreaking approach combining machine learning with cryptographic verification,” Dawn Song, Director of Berkeley RDI, commented. “We are excited to work together with Polyhedra to push the AI industry forward ensuring that the benefits brought by AI do not come at the expense of trust and safety.”</p><h4>The Future of Verifiable AI</h4><p>As we continue developing zkML alongside our Expander proving system, our vision is to make AI systems more transparent, accountable, and reliable. The synergy between zkML and blockchain technology unlocks opportunities for:</p><ul><li><strong>Secure AI Model Deployment</strong></li><li><strong>Verifiable Training Processes</strong></li><li><strong>Configurable Privacy Controls</strong></li><li><strong>Decentralized AI Ecosystems</strong></li><li><strong>Innovative AI Applications and Services</strong></li></ul><p>We are excited about the role zkML will play in shaping the next generation of AI applications, empowering trust and verifiability in ways that will redefine industry standards. This is just the beginning of what zkML can offer, and we look forward to sharing more as this technology evolves.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=283d03b990a5" width="1" height="1" alt=""><hr><p><a href="https://medium.com/polyhedra-network/berkeley-rdi-polyhedra-unveil-production-ready-zkml-4-years-after-pioneering-the-concept-283d03b990a5">Berkeley RDI &amp; Polyhedra Unveil Production-Ready zkML: 4 Years After Pioneering the Concept</a> was originally published in <a href="https://medium.com/polyhedra-network">Polyhedra Network</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Scaling Trust in the Age of Artificial Intelligence]]></title>
            <link>https://medium.com/@vreeland/scaling-trust-in-the-age-of-artificial-intelligence-3af459deee03?source=rss-b3599ef94fdc------2</link>
            <guid isPermaLink="false">https://medium.com/p/3af459deee03</guid>
            <dc:creator><![CDATA[Eric Vreeland]]></dc:creator>
            <pubDate>Tue, 19 Nov 2024 17:00:13 GMT</pubDate>
            <atom:updated>2024-11-25T20:18:32.992Z</atom:updated>
            <content:encoded><![CDATA[<p><em>First Practical Zero-Knowledge Proof System to Verify Transformer and Large Language Models</em></p><h3>Introduction</h3><p>Trust plays a crucial role in our everyday lives, shaping our interactions in ways we often take for granted. Imagine unlocking your phone with facial recognition, applying for a loan through an AI system, or receiving an AI-powered medical diagnosis. How can you be sure these AI systems are treating you fairly and protecting your privacy? As artificial intelligence increasingly shapes our daily lives, we face a critical challenge: building trust in AI systems while protecting sensitive information on all sides.</p><p>Consider a hospital using an AI system for cancer diagnosis. The hospital needs proof the AI performs consistently and accurately, but the AI provider can’t share proprietary model details. Meanwhile, patient privacy must be protected. Or picture an autonomous vehicle accident — how do we definitively determine whether the AI or human driver was at fault? Traditional approaches to AI transparency fall short in these scenarios, as they typically require exposing sensitive details that could compromise security, privacy, or intellectual property.</p><p>Zero-knowledge machine learning (zkML) offers a breakthrough solution to these real-world challenges. Several years ago, we first introduced the notion of zero-knowledge machine learning (zkML) in our seminal work “Zero Knowledge Proofs for Decision Tree Predictions and Accuracy” [1], which demonstrated how to construct zero-knowledge proofs for both predictions and accuracy of decision tree models. This innovative technology enables mathematical verification of AI systems while keeping sensitive information private. For example, a bank can prove its loan approval AI treats all applicants fairly without revealing its proprietary algorithms. Healthcare providers can verify AI diagnostic results while protecting both patient data and model details. And in the future, zkML could provide definitive proof of decision-making processes in accidents involving AI systems, clearly establishing responsibility between human and AI agents.</p><p>The applications of zkML extend across industries. Financial institutions can demonstrate regulatory compliance of AI trading systems without exposing strategic advantages. Employers can prove their AI recruitment tools are unbiased while protecting candidate data. Smart home devices can verify they protect user privacy while delivering personalized services. Government agencies can audit AI systems for fairness and safety without compromising security.</p><p>However, implementing zkML in practice faces significant challenges. Current systems require substantial computational resources, making real-time verification difficult for complex AI models. Integration with existing AI development workflows remains cumbersome, requiring specialized expertise. While research has shown zkML’s potential with various AI architectures — from decision trees to large language models — creating practical, user-friendly tools for widespread adoption remains a work in progress.</p><h3>Summary of Innovation</h3><p>The performance of our zkML technology represents a significant leap forward in the field of secure machine learning. By achieving a performance improvement of <strong>four orders of magnitude</strong> over previous approaches, we have drastically reduced the time and computational resources required for proof generation. This enhancement is exemplified by Expander’s ability to generate proofs for the <strong>Llama-3 model with 8 billion parameters</strong> in just minutes using a high-performance CPU. Such efficiency was previously unattainable and sets a new benchmark for what’s possible in zkML applications.</p><p>Moreover, this advancement marks the <strong>first time zkML has been successfully scaled to accommodate modern transformers of this size</strong>. Scaling zkML to work with large-scale, state-of-the-art models like Llama-3 not only demonstrates the robustness of our technology but also its applicability to real-world, high-performance machine learning tasks. This breakthrough enables secure and efficient deployment of complex models, ensuring that robust fraud prevention mechanisms can be implemented without sacrificing system performance or scalability.</p><p>Besides groundbreaking performance, Polyhedra has also developed zkPyTorch, a revolutionary compiler framework for zero-knowledge machine learning that seamlessly integrates with PyTorch. This groundbreaking system represents more than just a toolchain — it’s the first comprehensive compiler framework that enables direct translation of standard PyTorch models into zero-knowledge circuits. zkPyTorch dramatically simplifies zkML implementation through automated optimization and compilation, enhancing AI model security and privacy while ensuring compliance with standards. It sets a new benchmark for secure AI deployment in real-world applications and fundamentally transforms how developers can build verifiable AI systems.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*A1vbStXFIHvtcc-3Nr6OTw.jpeg" /></figure><p>As AI systems become more prevalent and sophisticated, the ability to verify their behavior while protecting sensitive information becomes increasingly crucial. zkML provides a mathematical foundation for building trustworthy AI systems that respect privacy, ensure consistency, and enable clear accountability.</p><p>Addressing these challenges requires more than just research — it demands a team capable of turning theory into practical solutions. At Polyhedra Network, we combine expertise in zero-knowledge proofs and AI with hands-on engineering experience to bridge this gap. Our work is carried out in close collaboration with researchers from the Berkeley Center for Responsible, Decentralized Intelligence (RDI), building on a shared legacy of innovation in ZK technology and decentralized systems.</p><p>Our work, such as verifying large language models like Llama-3 8B, showcases our ability to handle complex AI architectures and advance zkML technology. With a focus on usability and scalability, we develop solutions that bridge the gap between research and real-world applications. By integrating zkML into frameworks like PyTorch, we aim to make this technology accessible and practical for a wide range of industries.</p><h3>Zero-knowledge Machine Learning in a nutshell</h3><p>zkML leverages zero-knowledge proofs to enable verification of ML computations while preserving the privacy of models, input data, or both. Through this technique, one party can prove to another that they correctly conduct inference on a given model and a given input without revealing any sensitive information — whether that’s model parameters, input data, or intermediate computations.</p><p>Since our pioneering work [1], zkML has emerged as a fundamental approach for privacy-preserving machine learning applications, particularly in scenarios where both privacy and verifiability are essential requirements. A key functionality that zkML enables is <strong>non-interactive verification of inference</strong> — a service provider can generate a zero-knowledge proof demonstrating that a given output was legitimately produced by running the specified model on the input, whether it’s a decision tree, neural network, or other ML architecture.</p><p>Beyond verifiable inference, zkML’s capabilities extend to:</p><ul><li>AI Model Audit: Verifying a model’s performance on a specified test set through zkML, ensuring accuracy and compliance without revealing sensitive model or data details. This can be particularly useful for AI safety.</li><li>Training Data Origin Verification: Tracing and proving the lineage and authenticity of training data</li><li>Authenticated Data Labeling: Verifying the integrity of data labeling processes</li><li>Training Process Verification: Ensuring model training adheres to specified protocols and requirements.</li></ul><p>This comprehensive approach to verification addresses the full lifecycle of machine learning systems, from data preparation through model deployment and inference, establishing zkML as a cornerstone technology for trustworthy AI systems.</p><h3>Challenges and solutions</h3><p>The past four years have witnessed remarkable advancements in zkML technologies since our groundbreaking work in zkDT [1]. The field has expanded significantly, with subsequent innovations including zkCNN [2, 3, 12] and zkLLM [11], which extended zero-knowledge proof systems to neural networks and large language models respectively. Our pioneering contributions didn’t stop at theoretical foundations — we developed new, advanced distributed proving systems in [4,5,6,7] that have pushed the boundaries of what’s possible in zkML.</p><p>On the practical front, we’ve seen encouraging adoption of these ideas in real-world applications. Notable implementations include zero-knowledge proofs for nanoGPT (developed by the EZKL team [8]) and GPT-2 (developed by Modulus Labs [9]). These projects demonstrate the practical viability of zkML concepts, while also highlighting key challenges in the field.</p><p>However, significant challenges remain that our ongoing work aims to address:</p><ul><li><strong>Performance Limitations</strong>: Current proving systems like Halo2, when applied to models such as nanoGPT and GPT-2, achieve only <strong>sub-hundred</strong> parameters per second even on high-performance servers. This falls significantly short of practical requirements for real-world applications, where modern ML models contain billions of parameters. This performance gap represents a critical barrier to deploying zkML solutions at scale.</li><li><strong>Developer Accessibility</strong>: The current ecosystem requires developers to write customized circuits for ML models, demanding expertise in both cryptography and specialized programming languages. This steep learning curve presents a significant barrier to widespread adoption.</li></ul><p>These challenges underscore the importance of our continued work in advancing zkML technologies, particularly in developing more efficient proving systems and creating more developer-friendly tools and frameworks.</p><h3>Our Solutions</h3><h3>Expander</h3><p>We developed Expander, the world’s fastest ZK prover, implementing the next generation GKR protocol. With Expander we achieved an unprecedented 2.16 million Poseidon hashes per second on an AMD Ryzen 7950X3D and scales up to 16M hashes per second on 256-core server hardware [13]. This represents a transformative advancement for zkML implementations.</p><p>The performance breakthroughs are remarkable: we can prove VGG-16 with 15.2 million parameters in around 2 seconds on a single core CPU. We can also prove Llama-3 with 8 Billion parameters in around 2.7 hours on a single core CPU. Amortized, we can prove 0.8 million parameters per second, that is <strong>4 orders of magnitude faster</strong> than any number previously reported as shown in the table below.</p><p>And we didn’t stop here. We recognize that 166 minutes per 65 token inference is still far from ideal for practical applications. We are actively building hardware accelerations for our scheme [10] and developing a distributed prover infrastructure where multiple nodes can collaboratively work on the same proofs [6, 7]. Through these innovations, we are moving steadily toward our goal of <strong>real-time proving</strong> — where proof generation matches inference speed, effectively eliminating latency in the user experience.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3kw27bUcp44mkTHv12aHmQ.png" /></figure><h3>zkPyTorch</h3><p>In the ZK community, we have observed that developer accessibility remains a major obstacle to widespread adoption. Developers consistently express frustration about the need to learn new languages and understand sophisticated mathematics for zkML development. To address this challenge, we’ve developed a groundbreaking solution: a specialized compiler that works directly with PyTorch, eliminating the traditional barriers to zkML development.</p><p>Our compiler bridges the gap between AI and ZK by offering <strong>native PyTorch support</strong>, allowing developers to continue writing standard PyTorch code without learning new ZK-specific programming patterns. This seamlessly integrates with existing ML workflows while providing automated circuit optimization under the hood. The compiler intelligently translates PyTorch operations to ZK circuits and implements built-in optimizations for common ML patterns, ensuring efficient memory and computation layouts.</p><p>We implemented several proof system optimizations in our compiler. These focused primarily on faster computation verification for common mathematical operations. For example, we improved the speed of verifying Fast Fourier Transform calculations by reducing the number of required steps. We also made matrix multiplication checks more efficient through improved circuit design. These changes helped make our compiler’s verification process faster while maintaining correctness.</p><p>On the AI side, we utilize PyTorch alongside the XLA compiler to perform efficient and commonly used optimizations. Additionally, by leveraging the precomputed inference results, we significantly accelerate the verification process by treating the inference result as a prefill to the model. We also rely on the latest quantization result, which can significantly reduce the number of FP32 multiplication operations.</p><p>The end-to-end pipeline handles everything from automatic circuit generation to integration with our Expander proving system, culminating in a streamlined deployment process. This comprehensive approach dramatically reduces development time from months to days for complex models, eliminates the need for ZK expertise, and ensures code maintainability through standard PyTorch practices. Perhaps most importantly, our solution is future-proof — as ML models evolve, developers can easily update their implementations without learning new ZK-specific techniques.</p><h3>Future directions</h3><p>While our current achievements with Expander and zkPyTorch represent significant breakthroughs in zkML, we see several promising directions for further optimization and enhancement:</p><ol><li><strong>Smaller Finite Field Implementation</strong> Our current implementation uses BN254 curve, but migration to Mersenne-31 fields could potentially yield up to a 10x performance boost by reducing computational overhead in finite field operations.</li><li><strong>Multi-Machine Multi-GPU Support</strong> We are developing support for distributed computing environments leveraging multiple machines and GPUs, enabling parallel proof generation and efficient workload distribution for large-scale models.</li><li><strong>Extended Output Sequence Length</strong> While currently optimized for single-token output verification, we are developing optimizations to support efficient verification of longer output sequences, reducing per-token proving overhead and enabling practical applications requiring extended output generation.</li><li><strong>ZK-Friendly Quantization Techniques</strong> We are researching quantization methods specifically optimized for zero-knowledge proofs, focusing on minimizing arithmetic complexity in finite fields while maintaining model accuracy.</li><li><strong>Better compiler optimization </strong>Our compiler pipeline can be further optimized, potentially reducing both circuit size and proving time.</li></ol><p>These improvements will work synergistically with our existing optimizations to further reduce proving time and increase throughput. Our commitment to advancing zkML technology remains steadfast, and we welcome collaboration from the community as we work towards these</p><h3>Applications</h3><p>zkML applications span both Web2 and Web3 environments, each with distinct value propositions. In Web2 settings, zkML serves as a premium <strong>verification laye</strong>r for model providers, enabling two crucial trust guarantees. First, it provides proof to users that the correct model is being applied to the correct data, acting as a safeguard against both malicious service providers and potential model misbehavior. Second, as established in our zkDT paper [1], it offers a rigorous method to verify model accuracy by enabling providers to generate proofs of correct inference across large sets of input data. Together, these capabilities create a compelling value proposition that providers can offer as a premium service.</p><p>In Web3 environments, zkML serves a dual purpose of ensuring verifiability and privacy. On the verifiability front, we’ve pioneered the concept of <strong>Opportunistic zkML Proofs</strong>, implementing a challenge-based verification system where proofs are generated only when service providers are challenged. This approach offers numerous advantages: it reduces computational overhead during normal operation, creates strong fraud deterrence through economic incentives, lowers transaction costs when unchallenged, and enables better throughput for applications where immediate verification isn’t critical. The system provides flexible security, allowing applications to balance performance and security based on their specific needs.</p><p><strong>Privacy protection </strong>is important in systems that need to handle sensitive information<strong>.</strong> zkML makes it possible to keep data, models, and computations private while still allowing others to verify results. This ensures that both confidentiality and trust can be maintained across various applications.</p><p>In both worlds, our solution is designed with<strong> regulatory compliance</strong> as a primary consideration, implementing privacy as an opt-in feature. We’ve developed regulatory-first privacy controls that can be enabled based on specific requirements, with configurable settings to meet different regional and national regulations. The platform maintains a transparent mode by default when privacy features aren’t required, while providing clear tracking of privacy settings for regulatory reporting and audits. This flexible approach supports diverse use cases across sectors: from full transparency for public sector applications to HIPAA-compliant controls for healthcare, specialized settings for financial services, and configurable options for cross-border operations requiring compliance with multiple jurisdictional requirements.</p><h3>Looking Forward</h3><p>As we continue to develop zkML techniques alongside our Expander system, we’re committed to making AI systems more transparent, verifiable, and trustworthy, while respecting regulatory requirements and privacy needs. The zkML technology opens new possibilities for:</p><ul><li>Secure AI Model Deployment</li><li>Verifiable Training Processes</li><li>Configurable Privacy Controls</li><li>Decentralized AI Ecosystems</li><li>Novel AI Applications and Services</li></ul><p>We envision our solution as a foundational platform that will enable the next generation of verifiable AI applications. The examples shared here are just the beginning of what’s possible with this technology.</p><p>Stay tuned for more updates as we build the future of verifiable AI systems.</p><p><em>References:<br></em>[1] Zhang, J., Fang, Z., Zhang, Y., &amp; Song, D. (2020). Zero Knowledge Proofs for Decision Tree Predictions and Accuracy. In Proceedings of the 2020 ACM SIGSAC Conference on Computer and Communications Security (CCS ‘20).</p><p>[2] Feng, B., Qin, L., Zhang, Z., Ding, Y., &amp; Chu, S. (2021). ZEN: An Optimizing Compiler for Verifiable, Zero-Knowledge Neural Network Inferences. Cryptology ePrint Archive, Paper 2021/087. Available at:<a href="https://eprint.iacr.org/2021/087"> https://eprint.iacr.org/2021/087</a></p><p>[3] Liu, T., Xie, X., &amp; Zhang, Y. (2021). zkCNN: Zero Knowledge Proofs for Convolutional Neural Network Predictions and Accuracy. Cryptology ePrint Archive, Paper 2021/673. Available at:<a href="https://eprint.iacr.org/2021/673"> https://eprint.iacr.org/2021/673</a></p><p>[4] Xie, T., Zhang, J., Zhang, Y., Papamanthou, C., &amp; Song, D. (2019). Libra: Succinct Zero-Knowledge Proofs with Optimal Prover Computation. Cryptology ePrint Archive, Paper 2019/317. Available at:<a href="https://eprint.iacr.org/2019/317"> https://eprint.iacr.org/2019/317</a></p><p>[5] Zhang, J., Xie, T., Zhang, Y., &amp; Song, D. (2019). Transparent Polynomial Delegation and Its Applications to Zero Knowledge Proof. Cryptology ePrint Archive, Paper 2019/1482. Available at:<a href="https://eprint.iacr.org/2019/1482"> https://eprint.iacr.org/2019/1482</a></p><p>[6] Liu, T., Xie, T., Zhang, J., Song, D., &amp; Zhang, Y. (2023). Pianist: Scalable zkRollups via Fully Distributed Zero-Knowledge Proofs. Cryptology ePrint Archive, Paper 2023/1271. Available at:<a href="https://eprint.iacr.org/2023/1271"> https://eprint.iacr.org/2023/1271</a></p><p>[7] Xie, T., Zhang, J., Cheng, Z., Zhang, F., Zhang, Y., Jia, Y., Boneh, D., &amp; Song, D. (2022). zkBridge: Trustless Cross-chain Bridges Made Practical. In Proceedings of the 2022 ACM Conference on Computer and Communications Security (CCS ‘22).</p><p>[8] Modulus Labs. (2024). Chapter 14: The World’s 1st On-Chain LLM. Medium. Available at:<a href="https://medium.com/@ModulusLabs/chapter-14-the-worlds-1st-on-chain-llm-7e389189f85e"> https://medium.com/@ModulusLabs/chapter-14-the-worlds-1st-on-chain-llm-7e389189f85e</a></p><p>[9] EZKL. (2023). Honey I SNARKED the GPT. Available at:<a href="https://medium.com/@ModulusLabs/chapter-14-the-worlds-1st-on-chain-llm-7e389189f85e"> </a><a href="https://blog.ezkl.xyz/post/nanogpt/">https://blog.ezkl.xyz/post/nanogpt/</a></p><p>[10] Polyhedra Network. “Accelerating Ethereum’s Verge: Harnessing GPU Power for Zero-Knowledge Proofs.” GitHub Repository, PolyhedraZK/blogs, 2024.<a href="https://github.com/PolyhedraZK/blogs/blob/main/blogs/sumcheck_cuda.md"> https://github.com/PolyhedraZK/blogs/blob/main/blogs/sumcheck_cuda.md</a>.</p><p>[11] Sun, H., Li, J., &amp; Zhang, H. (2023). “zkLLM: Zero Knowledge Proofs for Large Language Models.”</p><p>[12] D. Kang, T. Hashimoto, I. Stoica, and Y. Sun, “Scaling up Trustless DNN Inference with Zero-Knowledge Proofs,” arXiv preprint, Oct. 2022. [Online]. Available: arXiv:2210.08674</p><p>[13] Expander performance. <a href="https://medium.com/polyhedra-network/expander-still-the-worlds-fastest-zk-prover-6aa5fd428609">https://medium.com/polyhedra-network/expander-still-the-worlds-fastest-zk-prover-6aa5fd428609</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3af459deee03" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Polyhedra Network and WSPN Join Forces to Enhance WUSD with Zero-Knowledge Technology]]></title>
            <link>https://medium.com/polyhedra-network/polyhedra-network-and-wspn-join-forces-to-enhance-wusd-with-zero-knowledge-technology-cbb28309538b?source=rss-b3599ef94fdc------2</link>
            <guid isPermaLink="false">https://medium.com/p/cbb28309538b</guid>
            <category><![CDATA[interoperability]]></category>
            <category><![CDATA[polyhedra-network]]></category>
            <category><![CDATA[stablecoin-cryptocurrency]]></category>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[proof-of-reserve]]></category>
            <dc:creator><![CDATA[Eric Vreeland]]></dc:creator>
            <pubDate>Tue, 08 Oct 2024 19:12:22 GMT</pubDate>
            <atom:updated>2024-10-09T01:59:44.088Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZofFiPE8Zoh9_qPeN-aQXQ.jpeg" /></figure><p>Polyhedra Network is excited to partner with Worldwide Stablecoin Payment Network (WSPN) to incorporate zero-knowledge technology into the future of digital payments. Leveraging Polyhedra’s expertise in cutting-edge cryptographic techniques, this partnership aims to drive innovation, speed, and security for blockchain technology used by international payment systems.</p><p>WSPN provides transparent and efficient solutions for payment technology, with a dedication to shaping seamless experiences at the frontier of financial inclusion. In particular, WSPN pioneers the stablecoin WUSD, which is pegged 1:1 to the U.S. Dollar and aims for worldwide payments and round-the-clock financial markets. WUSD is built to bridge the gap between DeFi and TradFi, uniquely poised to transact across borders and mitigate volatility associated with other riskier assets. WSPN has taken a proactive approach with regards to compliance for WUSD, partnering with security firms like Certik to ensure smart contract integrity.</p><p>There are three areas the joint teams have identified for collaboration:</p><p><strong>Interoperability with zkBridge </strong><br>Bringing zkBridge to WSPN’s infrastructure will help increase interoperability without relying on external trust assumptions. zkBridge allows different networks to verify data in a secure manner, and employs the use of best-in-class provers to significantly improve proof generation. zkBridge also reduces on-chain verification costs, and will allow for transference of cross-chain assets and liquidity. Together, zkBridge and WSPN hope to benefit the overall functionality, data validity, and collaboration of the WUSD system.</p><p><strong>ZK-powered Proof of Reserve</strong></p><p>Polyhedra will build upon WSPN’s current Proof of Reserve process by using a ZK PoR mechanism. Polyhedra PoR will help demonstrate and ensure that assets exist 1–1 to back the WUSD in circulation. This verifies that every dollar in WUSD is backed by its equivalent tangible asset in the real world. As a result, users can have more trust in the stability of their assets.</p><p><strong>Pioneering Private Transactions</strong></p><p>Historically, the transparency of blockchains has been a sore point for financial institutions that require privacy and regulatory measures such as selective disclosure. Polyhedra has developed a proprietary offering that employs conditional de-anonymization for compliant and efficient transactions. In the context of WUSD, this solution will empower users and institutions with the ability to obfuscate their transaction details.</p><p>All three of these solutions are supported by <a href="https://medium.com/polyhedra-network/introducing-expander-the-fastest-gkr-proof-system-to-date-bdd07d05c23e">Expander, the fastest ZK proof system to date. </a>Invented by the cryptography team at Polyhedra, Expander uses innovative polynomial commitment schemes integrated with the GKR protocol to minimize computational overhead, reduce memory usage, and process high-performance calculations. Expander’s contribution to these products and this collaboration means WSPN can set a new industry standard for stablecoin scalability and security.</p><p>Together, WSPN and Polyhedra are excited about how this partnership will expand ecosystems and drive value for users and the community. Increasing usability of these tools will help to bridge the gap between traditional finance and the digital economy.</p><p><strong>About Polyhedra Network</strong></p><p>Polyhedra Network is building the next generation of secure, interoperable, and scalable blockchain infrastructure with advanced zero-knowledge-proof technology. The team has developed and deployed several next-generation ZK-SNARK protocols resulting in performance improvements orders of magnitude faster than existing solutions. Their interoperability solution, zkBridge, connects over 25 blockchains and proves the validity of cross-chain messages utilizing zero-knowledge proofs. Developers building with Polyhedra Network can create secure, scalable, interoperable applications without any additional trust assumptions.</p><p><a href="https://polyhedra.network/?ref=blog.polyhedra.network">Website</a> | <a href="https://twitter.com/PolyhedraZK?ref=blog.polyhedra.network">X (formerly Twitter</a>) | <a href="https://discord.gg/polyhedra-network?ref=blog.polyhedra.network">Discord</a></p><p><strong>About WSPN</strong></p><p>WSPN is a leading provider of next-generation stablecoin infrastructure, committed to building a more secure, efficient, and transparent payment solution for the global economy. Their flagship product, WUSD stablecoin, is pegged 1:1 to the U.S. Dollar and aims to optimize secure digital payments for Web3 users. WSPN ‘s Stablecoin 2.0 approach prioritizes user-centricity, community governance, and accessibility, paving the way for widespread stablecoin adoption.</p><p>Learn more:<a href="http://www.wspn.io/"> www.wspn.io</a> | <a href="https://x.com/wspnpayment">X</a> |<a href="https://www.linkedin.com/company/wspn/"> LinkedIn</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cbb28309538b" width="1" height="1" alt=""><hr><p><a href="https://medium.com/polyhedra-network/polyhedra-network-and-wspn-join-forces-to-enhance-wusd-with-zero-knowledge-technology-cbb28309538b">Polyhedra Network and WSPN Join Forces to Enhance WUSD with Zero-Knowledge Technology</a> was originally published in <a href="https://medium.com/polyhedra-network">Polyhedra Network</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The First Proof Arena Benchmarks Are Live!]]></title>
            <link>https://medium.com/polyhedra-network/the-first-proof-arena-benchmarks-are-live-56f482712977?source=rss-b3599ef94fdc------2</link>
            <guid isPermaLink="false">https://medium.com/p/56f482712977</guid>
            <category><![CDATA[web3]]></category>
            <category><![CDATA[polyhedra-network]]></category>
            <category><![CDATA[keccak256]]></category>
            <category><![CDATA[cryptography]]></category>
            <category><![CDATA[zero-knowledge-proofs]]></category>
            <dc:creator><![CDATA[Eric Vreeland]]></dc:creator>
            <pubDate>Mon, 30 Sep 2024 15:12:02 GMT</pubDate>
            <atom:updated>2024-10-02T16:30:38.342Z</atom:updated>
            <content:encoded><![CDATA[<p>We are thrilled to announce that the first set of benchmarks in <a href="https://proofarena.org/">Proof Arena</a> is now live! This marks a significant milestone in our journey to revolutionize the world of zero-knowledge proofs (ZKPs) and empower developers to build more secure and efficient ZK applications.</p><h3>What is Proof Arena?</h3><p>Proof Arena is an innovative benchmarking platform designed to evaluate and compare different zero-knowledge provers. It provides a structured environment where developers can submit their prover implementations, which are then rigorously tested against a series of predefined problems. Problems are specific computational challenges that applications frequently encounter and utilize zero-knowledge to solve. The platform measures key performance metrics, including setup time, witness generation time, proof generation time, verification time, peak memory usage, and proof size.</p><p>The importance of each factor varies based on the application:</p><ul><li>For <strong>real-time applications</strong>, proof generation time may be the most critical because processing speed heavily impacts user experience.</li><li>In <strong>high-throughput systems</strong>, verification time could take precedence due to the need for quick verification of numerous proofs.</li><li>For <strong>resource-constrained environments</strong>, peak memory usage is paramount as developers need to ensure limited resources are being effectively allocated.</li><li>In <strong>data-intensive applications</strong>, proof size may be the key consideration to ensure scalability.</li></ul><p>By offering a transparent and competitive space for prover evaluation, Proof Arena aims to foster collaboration and innovation within the zero-knowledge community. Developers can gain insights into the performance of their solutions, identify areas for improvement, and contribute to the advancement of ZKP technology.</p><h3>Proof Arena Adoption</h3><p>Over the past two months, the response to the release of Proof Arena has been tremendous. Within the web3 space, the teams from Polygon, Linea, and StarkWare have contributed to the platform to ensure that their proof systems, Plonky3, GNARK, and Stwo respectively, are represented fairly and accurately.</p><p>In academia, Proof Arena has drawn the attention of some of the foremost experts in zero-knowledge cryptography including Professor Dawn Song, co-Director of the UC Berkeley Center on Responsible Decentralized Intelligence, and Professor Dan Boneh, head of the applied cryptography group at Stanford.</p><p>“Proof Arena provides a long-awaited tool to compare different ZK provers on different tasks,” Professor Boneh shared. “Proof Arena updates every time a new prover version is released, and moreover, new provers can be easily added. This keeps the comparison current and up-to-date with the latest developments in the field.”</p><p><strong>Results from the First Live Problem</strong></p><p>We are excited to share the results from the initial set of problems that were evaluated with Proof Arena. The data collected provides valuable insights into the performance of various provers, showcasing their strengths and weaknesses. Determining the most important factors to consider depends on the specific application and context in which the proof system is being used. Here’s a breakdown of each factor’s significance:</p><ol><li><strong>Proof Generation Time</strong>: This is crucial in scenarios where real-time performance is essential, such as in blockchain transactions or interactive applications. If the proof generation time is too long, it can lead to delays in processing transactions or responses, negatively impacting user experience.</li><li><strong>Verification Time</strong>: This is particularly important in systems where many proofs need to be verified quickly, such as in batch processing or when multiple users are interacting with a system simultaneously. Faster verification times can enhance the overall throughput of the system.</li><li><strong>Peak Memory</strong>: In environments with limited resources, such as IoT devices or mobile applications, peak memory usage can be a critical factor. High memory consumption can lead to performance degradation or even system crashes, making it essential to optimize memory usage.</li><li><strong>Proof Size</strong>: Smaller proof sizes are beneficial for storage and transmission efficiency, especially in bandwidth-constrained environments. In blockchain applications, smaller proofs can reduce the amount of data that needs to be stored and transmitted, which is vital for scalability.</li></ol><p>In general, the performance of these four metrics will take precedence over setup time and witness generation time.</p><p><strong>Deep Dive: Keccak256 Hashes</strong></p><p>Keccak256 hashes, a member of the SHA-3 family, play a pivotal role in blockchain development due to their robust security and efficiency. As a cryptographic hash function, Keccak256 generates a fixed-size 256-bit output from input data of any length, ensuring that even the slightest change in the input results in a drastically different hash. This property is crucial for maintaining data integrity and authenticity on the blockchain, as it helps prevent tampering and ensures that transactions are securely linked to their corresponding data. Additionally, Keccak256’s resistance to collision and pre-image attacks makes it a reliable choice for various blockchain applications, including smart contracts and zero-knowledge proofs. By leveraging Keccak256, developers can enhance the security of their decentralized applications, ensuring that sensitive information remains protected while still enabling efficient verification processes.</p><p>Given the significance of Keccak256 hashes in blockchain development, it is fitting that this is the first problem with performance metrics generated across a wide variety of provers. As you can see from the results below, the <strong>Expander</strong> prover is significantly faster in terms of proof generation time and peak memory, <strong>Plonky3</strong> achieves stellar performance in terms of setup and verification time, and <strong>GNARK </strong>has the smallest proof size.</p><p>In use cases like bridging assets, where near real-time execution is critical, proof generation time is the primary consideration, which is why we developed Expander while building zkBridge.</p><p>The complete results for this problem as well as others can be viewed on the <a href="https://proofarena.org/problems/2">Proof Arena website here</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ASJMVyGnrE3NUfjCez61ng.jpeg" /></figure><p>Diving deeper into the data we can see that the range for proof generation time is quite large. Expander leads the pack with about 16,700 proofs per second followed by Plonky3 at 1,368 and then GNARK and Halo2 at 4.5 and less than 1 respectively.</p><p>On the other hand, the spread between verification times is smaller. Expander verifies about 928 proofs per second whereas Plonky3 verifies 5,413 proofs per second.</p><p>While both proof generation time and verification time are important, there are specific contexts where proof generation time should take precedence, especially given the variance in proof generation performance. Real-time applications, high-volume transaction environments, and resource-constrained systems should seriously consider proof generation time when selecting the best prover for their use case.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*C_Anl4bKMNm2rJZr" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Xa-acMXc6BF9Je74" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*-ZH_2vg8QomVaa-z" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*gR5GKKotOEfr6hVD" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ViXmJKSvVyv6b_Oi" /></figure><h3>Why This is Significant for Teams Utilizing Zero-Knowledge</h3><p>The launch of Proof Arena and the results from the first live problems hold significant implications for teams leveraging zero-knowledge technology:</p><ol><li><strong>Informed Decision-Making</strong>: With access to comprehensive performance data, developers can make informed decisions about which prover to use for their applications. This empowers teams to optimize their solutions based on real-world performance metrics.</li><li><strong>Enhanced Collaboration</strong>: Proof Arena fosters a collaborative environment where developers can share insights, learn from each other, and collectively advance the state of zero-knowledge proofs. This community-driven approach accelerates innovation and encourages best practices.</li><li><strong>Improved User Experience</strong>: By selecting the most efficient provers, teams can enhance the user experience of their dApps. Faster verification times and smaller proof sizes lead to quicker transactions and lower costs, ultimately benefiting end-users.</li><li><strong>Driving Adoption</strong>: As more developers engage with Proof Arena and contribute their solutions, we anticipate a surge in the adoption of zero-knowledge technology across various industries. This will pave the way for more secure, scalable, and private applications, aligning with the core values of the Web3 movement.</li></ol><h3>Join Us on This Journey!</h3><p>We invite all developers and teams interested in zero-knowledge proofs to explore Proof Arena, submit their provers, and participate in this exciting benchmarking initiative. Together, we can push the boundaries of what is possible with zero-knowledge technology and create a more secure and decentralized future.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=56f482712977" width="1" height="1" alt=""><hr><p><a href="https://medium.com/polyhedra-network/the-first-proof-arena-benchmarks-are-live-56f482712977">The First Proof Arena Benchmarks Are Live!</a> was originally published in <a href="https://medium.com/polyhedra-network">Polyhedra Network</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>