Author: Whynona & Jake Koch-Gallup, Messari Research Analyst; Translation: Golden Finance xiaozou

1 , abstract

Nillion has reached cooperation with companies such as Virtuals, NEAR, Aptos, Arbitrum, Ritual, io.net and Meta.

A suite of application tools including nilAI, nilVM, nilDB and nilChain provides developers with the ability to create privacy-preserving applications across AI, healthcare and DeFi resources required by the program.

The network leverages privacy-enhancing technology (PET) orchestration—including multi-party computation (MPC), homomorphic encryption, and zero-knowledge proof techniques—to promote its decentralization Secure data computing and storage in centralized infrastructure.

Nillion’s validator program has covered approximately 500,000 validators, who actively contribute to the network and jointly process approximately 195 million Secrets, protecting Approximately 1,050 GB of data is safe.

2. Introduction

For a long time, processing high-value data (such as passwords, personalization artificial intelligence, healthcare information, and biometrics) are both unsafe and inefficient. While encryption secures stored data, it requires computations to be decrypted and then re-encrypted, introducing vulnerabilities and latency risks. Although blockchain technology decentralizes transactions and data management, it does not inherently solve the challenge of secure computing of encrypted data. This constraint limits the types of applications that can be safely built in Web3.

Nillion addresses these limitations by enabling data transmission, storage, and computation without decryption, ensuring that sensitive information remains private and secure throughout its lifecycle. This approach, called Blind Compute, decentralizes trust and expands decentralized network use cases into previously untouched “white space” areas, such as private AI agents, private LLM inference, and other secure Data demanding industries. By using advanced privacy-enhancing technologies (PET) such as multi-party computation (MPC) and fully homomorphic encryption (FHE)and Trusted Execution Environment (TEE), Nillion allows data to remain encrypted during computation.

3. Background

Nillion Network was established in 2021 and provides an innovative approach to handle private data across distributed systems without compromising security or efficiency. Powered by application frameworks such as nilVM, nilDB, nilAI and nilChain, Nillion provides developers with the tools to build privacy-preserving applications across areas such as AI, DeFi and data storage.

The project team includes former Hedera SPV partner and Goldman Sachs banker Alex Page (CEO); Hedera co-founder and Reserve founding chief marketing officer Andrew Masanto ( Chief Strategy Officer); Indiegogo founder Slava Rubin (Chief Brand Officer); Dr. Miguel de Vega (Chief Scientist), who holds more than 30 data optimization patents; Uber founding engineer Conrad Whelan (Founding Chief Technology Officer); Nike's former innovation director Mark McDermott (chief operating officer); Hedera’s early senior associate partner and former UBS and Rothschild banker Andrew Yeoh (chief marketing officer), among others.

The team has raised $50 million in private financing since its inception from investors including Hack VC, Hashkey Capital, Distributed Global and Maelstrom.

4. Technology

Nillion Network is a decentralized infrastructure designed to support Privacy-preserving secure operations for high-value data. It consists of two core layers: a coordination layer that manages governance and payments, and a Petnet that handles computation and storage. Nillion’s multi-party computation (MPC) protocol is at the core of the network’s capabilities, enabling private data computation without exposing personal input information. Nillion’s ecosystem is powered by a suite of application tools (namely nilAI, nilVM, nilDB, and nilChain) that enable developers to build privacy-focused applications.

(1)Nillion Network

Nillion Network is a decentralized infrastructure designed to support private , high-value data storage and computing. Scalability is achieved through clustering, which configures clusters of nodes to meet specific performance, security, and cost needs. Unlike traditional blockchains, the operation of the network does not rely on global shared state, supporting vertical scalability (achieved by upgrading individual nodes or clusters) and horizontal scalability (achieved by adding new nodes or clusters), to distribute workload efficiently. The following is the contribution of each layer of the network (i.e. coordination layer and Petnet) to the network architecture.

Coordination layer

The coordination layer of Nillion Network is called nilChain and is responsible for: managing rewards , payments, cryptoeconomic security, and inter-cluster coordination of networks.

Specifically, it coordinates the payment of storage operations and network-blind computation without directly handling the computation. The coordination layer is built with the Cosmos SDK and supports IBC for interoperability; however, given the network's core focus on storage and compute, it does not currently support smart contract execution. While directly accessible via Keplr or Leap wallets, applications created on the collaborative blockchain (detailed further in the Key Projects section below) will be completely abstracted. nilChain has been running on the test network since June 2024.

Petnet

Petnet (orchestration layer) is designed to integrate functions such as multi-party computation (MPC), Encryption technologies such as fully homomorphic encryption (FHE) and zero-knowledge proof (ZKP) to enable privacy-preserving computing and data management. This integration is achieved through two key components: the compiler and the computational network. Specifically, the compiler simplifies the use of privacy-enhancing technologies (PETs) by performing different levels of abstraction when computing networks perform secure computations and manage encrypted data.

This approach is deployed on the Nillion Network via the Nada language compiler and nilVM, with elements at all four different abstraction levels already in development. The four abstraction levels are as follows:

-Each PET protocol runs independently in its own blind module, similar to an isolated black box. There is no built-in unified interface or abstract processing, all orchestration occurs on the client side; therefore, developers can use the API to Perform specific tasks but cannot integrate or customize them

- Various blind modules are integrated into each SDK, providing developers with a simple and unified way to manage multiple PET protocols without the need for cryptographic expertise. Although these modules are not fully optimized, they are currently not Relying on a single PET protocol, but a seamless, ready-to-use combination of PET protocols is available

-Blind modules start supporting multiple PET protocols in a single blind module. This provides developers with the ability to make various trade-offs between performance and security - further simplifying the process for developers who lack expertise Encryption Decisions

- Blind modules are deployed on loosely independent networks, called clusters, and are managed by NilChain. As the Nillion blind computer matures, the same blind module can be replicated in multiple clusters, each with different configurations. These configurations depend on various factors ( e.g., number of nodes, node location, reputation, hardware specifications, security thresholds). This versatility allows developers to use the same functionality in different cluster setups, allowing targeting of specific needs (i.e. security, cost). , hardware, compliance, etc.) custom solutions

Nillion's PET technology is introduced in stages - each stage passing through the four levels of abstraction mentioned above. Phase 1 (i.e. HE, LSSS MPC) and Phase 2 (i.e. DWT+LSSS, TEE) are further in the abstraction process and have been integrated into Nillion Network. In stage 3 technologies (i.e., FHE-MPC, DWT+TEE, public computing, ZKP), FHE-MPC has begun to advance at the abstract level.

Working principle

The following is the detailed operation of each component of the Nillion network:

- Users/developers interact with front-end applications built with JavaScript or Python clients to submit data for storage or initiate blind calculation requests.

-Applications that use JavaScript clients to interact with Petnet for secure computing and encrypted data management. Python client-based applications interact with the coordination layer for payments, forwarding, and multi-chain communication. The coordination layer processes payments using the appropriate blockchain native gas token or NIL token.

-After the coordination layer processes the request, it forwards the calculation task to Petnet that supports PET.

-Petnet processes data by using PET techniques like linear secret sharing schemes, garbled circuits, and/or homomorphic encryption, depending on the task requirements. These calculations are performed across a cluster of nodes. Each node in Petnet only manages a fragment (share) of encrypted data.

-The node performs the specified calculation (for example, addition, multiplication, or safe comparison) on the masked data and generates partial output data.

-Petnet aggregates these partial outputs to generate the final calculation results in a secure and private manner.

-The final result path is as follows: If using a JavaScript client, Petnet sends the results directly to the application for user/developer access. If using a Python client, the coordination layer will retrieve the results from Petnet and send them to the application or relevant blockchain for further use.

-For blockchain integration use cases, the coordination layer delivers results to the original smart contract or decentralized application, supporting multi-chain functionality without requiring users to download new wallet.

(2) MPC protocol for complex operations

Multi-party computation (MPC) is A subfield of cryptography that enables individuals to collaboratively compute results from their combined data without revealing their input data. Nillion developed an MPC protocol called Curl, which is based on the Linear Secret Sharing Scheme (LSSS) but extends the performance to efficiently handle complex operations such as division, square roots, trigonometric functions, and logarithms. This makes Curl highly scalable and ideal for real-world problems such as privacy-preserving AI agents whose outputs do not scale linearly with the input. Curl follows a two-stage workflow structure:

Phase 1 (preprocessing to create shares): This phase generates random shares and assigns them to participants (computing entities) before processing the actual data using MPC technology. It is worth noting that the preprocessing stage operates independently of the input values, relying only on the number of inputs to create the appropriate number of shares before the calculation occurs. It can be thought of as an abstraction layer - creating placeholders ahead of time that will later be combined with the actual input data provided by the user in stage 2.

Phase 2 (Efficient Computation of Complex Operations): This computation phase involves the actual computation of private inputs in three steps: input, evaluation, and output. Input: Each party distributes its input share among participants to ensure information theoretic security (ITS). Each participant receives a share for each value entered, thus maintaining confidentiality throughout the process. Evaluation: Parties use Nillion's Curl protocol to efficiently compute complex operations on input shares. Output: Display and summarize local calculation results to produce a final result.

(3) Application tools

Application tools developed based on Nillion Network (nilVM, nilDB , nilAI, nada integration package) provides developers with a modular framework and practicality to quickly build privacy-preserving high-value data applications.

nilAI

nilAI is Nillion’s privacy technology suite focusing on artificial intelligence (i.e. AIVM, nada-AI and nilTEE). Here’s how they work:

- AI Virtual Machine (AIVM): a secure AI inference platform developed based on Nillion’s MPC technology and Meta’s CrypTen framework . It uses the discrete wavelet transform (DWT) developed with the Meta AI research team to accelerate inference. AIVM ensures private deep learning model inference and deployment by keeping individual nodes immune to user prompts and model output, thereby maintaining data confidentiality.

- nada-AI: nilVM library designed for AI applications, providing a PyTorch-like interface for running small models (e.g., neural networks, volumes Neural network, linear regression, etc.). Developers can also use theGoogle Colab notebooks jumpstart their projects.

- nilTEE: A solution for running large language models (llm) during high-performance inference using the Trusted Execution Environment (TEE). Nillion recommends limiting the use of TEE to inference time, not long-term data storage. Research is ongoing to enhance nilTEE and AIVM by splitting the inference setup to further improve security and performance.

nilVM, Nada and its code base

nilVM is a tool that allows developers to create The program's virtual machine. Programs are written in Nada (Nillion’s open source Python-based DSL) and developed using the Nillion SDK. Nada also includes nada-ai (similar to PyTorch and scikit-learn), nada-numpy, nada-data and nada-test code libraries to simplify program development. Developers can integrate nilVM into their applications using Python, Typescript or CLI clients and leverage the storage API for secure data storage and retrieval on the Nillion Network. Examples include federated learning programs, community development projects, and interactive demonstration use cases.

nilDB

nilDB is a cryptographically distributed NoSQL database designed for privacy-preserving data storage and computation. Unlike regular NoSQL databases, nilDB eliminates dependence on a central authority by distributing encrypted data as a secret share across multiple nodes. Additionally, data owners can grant others access to run SQL-like queries, calculations, and privacy-preserving aggregations on the stored data.

Here's how it works:

-Users encrypt sensitivity locally on their own devices data.

-Users securely upload encrypted data via a front-end application built on Nillion. The application securely uploads encrypted data to nilDB via an integrated backend RESTful API.

-The encrypted data is split into secret shares using Nillion’s MPC protocol and distributed across a cluster of nodes in the nilDB network. It is worth noting that no single node holds the complete dataset.

-The user provides explicit consent to the use or query of specific data, which authorization can be revoked at any time through the application.

- Authorized entities (such as companies or third parties) submit SQL-like query requests (such as lookups, range filters, or aggregate calculations) through Nillion's RESTful API ).

-Nodes in a nilDB cluster work together to perform calculations on encrypted data without disclosing sensitive information.

-Generate query results such as averages, sums, or filtered data sets while maintaining data confidentiality.

-Only the final query results are returned to the user who initiated the request through the RESTful API.

Nada integration package

Nada language includes various integration packages, including nada-ai, nada-numpy and nada-test, their use cases are as follows:

- nada-numpy: A constrained adaptation of NumPy tailored for Nada DSL. Compared with regular NumPy, nada-numpy allows efficient operations on array structures with strong type requirements for data types, ensuring compatibility with the strong type features of MPC.

- nada-test: The testing framework for Nada programs supports dynamic generation of tests at runtime. Developers can use Python to write test cases, integrate the framework into pytest workflows, and define flexible input and output specifications.

Other tools (e.g., Nada DSL, Nada Sandbox, etc.) and SDK can be found on GitHub.

(4) Research paper

Nillion's project team collaborated with multiple researchers and published eight papers detailing various aspects of the protocol and its applications.

Nillion: A Secure Processing Layer for Web3: An original vision document outlining the potential for Nillion and its applications in a decentralized ecosystem.

Evaluation of Arithmetic Sum-of-Products Expressions in Linear Secret Sharing Schemes with a Non-Interactive Computation Phase: Explore the Nillion MPC protocol for secure and efficient nonlinear computation

Curl: Private LLMs through Wavelet- Encoded Look-Up Tables: Propose the Curl framework (a privacy-preserving inference framework for LLM), which utilizes wavelet encoded look-up tables to reduce communication overhead and improve efficiency.

Technical Report on Secure Truncation with Applications to LLM Quantization: Explore secure truncation techniques in MPC environments based on Linear Secret Sharing Scheme (LSSS) for optimizing calculations in LLM.

More efficient comparison protocols for MPC: Enhance the efficiency of safety comparison in LSSS-based MPC systems

Technical Report on Threshold ECDSA in the Preprocessing. Setup: Detailed introduction to the preprocessing method of Threshold ECDSA, a distributed encryption system that securely manages and uses multi-party private keys

Technical Report on. Decentralized Multifactor Authentication: IntroductionIntroducing a decentralized framework for improving secure authentication processes.

Ripple: Accelerating Programmable Bootstraps for FHE with Wavelet Approximations: Detailed description of Ripple, a framework for compressing lookup tables using discrete wavelet transform (DWT), reducing identical Boot computation cost in state-of-the-art encryption (FHE).

5. NIL token

(1) Token function

Payment for computations across the Petnet and coordination layers Services, data storage, AI inference and transaction fees. Specifically, developers use NIL to access Nillion’s privacy-preserving computing services for their applications.

Stake and support network security and receive rewards. Validators bind NIL to verify transactions and calculations, thereby protecting the security of the coordination layer. Petnet nodes stake NIL to increase the security of their cluster and attract developers and applications.

Participate in decentralized governance by making suggestions and voting on various network decisions (e.g., protocol upgrades, resource allocation, and community grant programs).

(2) Governance

Governance decisions are made through the on-chain voting mechanism. Specifically, any NIL token holder can propose updates to the network, provided they meet the minimum token deposit requirements. Proposals may also be submitted by community boards or working groups established through previous governance actions.

Voting rights extend to key decisions such as:

Introducing new features or updates.

Allocate reward pools for donation grant programs, developer incentives, and community-driven projects.

Adjust network pricing, validator requirements, or delegation limits.

Modify governance structures such as quorum requirements or proposal thresholds.

Expand interoperability, form strategic partnerships, or deploy transparency and audit mechanisms.

Voting rights are proportional to the amount of NIL pledged, and pledgers delegate their voting rights to validators while retaining their own ability to vote on proposals.

6. Nillion ecosystem status

(1) Key areas

Key areas that highlight the various benefits of Nillion include:

Artificial Intelligence: Processing data and inference without disclosing sensitive information, bridging the gap between secure local AI processing and the scalability of centralized, non-proprietary AI systems. Personalized agents: AI agents can store, compute and process private data. Private model inference: AI models can securely handle private data, minimize the risk of information disclosure to third parties, and enable private LLM. Private knowledge base and search: Data can be stored in encrypted form while still enabling search capabilities for AI agents and other AI use cases.

Data ownership: Nillion's cryptographic infrastructure supports secure data marketplaces by allowing users to control and sell their data to buyers.

Blockchain: Nillion allows blockchain applications to send blind storage and computing requests to the Nillion network, complementing the public data capabilities of the blockchain. It also supports on-chain settlement by allowing applications to decrypt relevant data on the blockchain.

Healthcare: Nillion supports privacy-preserving analysis of healthcare data across institutions and users.

DePIN: The DePIN project integrates with Nillion to securely store and process sensitive operational data.

(2) Key projects

Key projects highlighting the various benefits of Nillion include:

Virtuals Protocol: an AI agent commons platform that developed a multi-modal AI agent library and Using Nillion allows private training and inference of its artificial intelligence models to build personalized artificial intelligence agents.

Aptos/NEAR/Arbitrum/Sei: Integrate L1 and L2 of blind data storage and calculation to enhance data processing in smart contracts.

Ritual: This artificial intelligence platform builds a decentralized artificial intelligence reasoning network and integrates Nillion on its backend for private reasoning.

Zap: This is a data platform that aggregates user data into Nillion's decentralized data pool through blind computing and zero-knowledge transport layer security (zkTLS ) provides security insights.

Reclaim Protocol: This is a zkTLS infrastructure platform that allows users to prove identity and reputation in a trustless manner through an off-chain platform and uses Nillion as a generated proof storage and processing platform.

Healthblocks: a fitness app that uses Nillion to maintain user ownership and control of their data while allowing third parties to do so without disclosing personal details Generate analytical insights.

MonadicDNA: A genomics platform that uses Nillion to keep data encrypted throughout its lifecycle, providing an alternative to centralized providers such as 23andMe choose.

(3) Nucleus Builders plan

The Nucleus Builders plan will be launched in January 2024. Support developers to create decentralized privacy-preserving applications using Nillion’s technology. Participants have access to the Nillion SDK, development tools such as nilVM, nilDB, and nilAI, as well as financial and technical resources to accelerate overall development. Since the program’s inception, Nucleus has attracted nearly 50 participants from more than 10 verticals, ecosystem projects have raised a total of more than $100 million.

(4) Network indicators

"Nillion verification announced on August 27, 2024 The "Verifier Program" allows participants to support the integrity and functionality of the network by uploading "Secrets", which are fragments of sensitive data of various categories, or setting up validators to challenge secrets and protect network security, ensuring network data sharing Maintain reliability and integrity.

Participation in the validator program ends on December 11, 2024, and approximately 500,000 validators have actively participated in the network, collectively processing approximately 195 million secrets And ensured the security of approximately 1,050 GB of data.

7. Roadmap

"Nillion Roadmap" released on May 31, 2024 "Contains four key stages:

Phase 1—Genesis Sprint (this phase has been completed): This phase established the underlying coordination layer during the launch of the testnet, tested core functions (such as Keplr wallet creation, token transfer, staking and governance), and provided The developers provided the Nillion SDK with telemetry for early application development, as well as load testing to evaluate transaction throughput and network scalability.

Phase 2 - Catalyst Aggregation (this phase is in progress): This phase integrates Petnet with the coordination layer and recruits external nodes to achieve complete decentralization. Introduce blind applications for secure data processing, support cross-chain functions, and expand Nillion into a multi-chain ecosystem.

Phase 3 - Strengthening: This phase will complete the mainnet launch and token generation event (TGE), run external nodes, and achieve real-world interaction through blind calculations , as well as verifying previously built applications on the network under real-time conditions.

Phase 4 - Multi-cluster future: This phase will focus on horizontal scaling by adding public node clusters, increasing computing power, optimizing the network for specific market applications, and Achieve scalability while ensuring security and privacy.

8. Conclusion

Nillion is a decentralized infrastructure designed to handle high-value, privacy-sensitive data in a variety of applications, from AI agents to private DeFi. Combined with advanced PET technologies (e.g., MPC, FHE, TEE), Nillion expands the usability of decentralized networks and the possibilities of decentralized applications. Nillion's architecture - coordination layer and Petnet - supports scalability through a clustering approach while ensuring data confidentiality and decentralized trust.

The Nillion ecosystem will continue to expand, with landmark milestones including: the Nucleus Builders program, which supports approximately 50 projects across multiple verticals; approximately 500,000 Validators participated, collectively processing approximately 195 million secrets and securing approximately 1050 GB of data. Collaborations with Virtuals, NEAR, Meta, and Aptos, along with ongoing roadmap progress for mainnet launch and multi-cluster scalability, highlight Nillion’s efforts to advance privacy-focused data management and secure computing.