System Architecture

The Privasea network embodies a sophisticated system architecture, meticulously designed to address the pressing challenges of privacy-preserving AI computations. Rooted in a robust design philosophy, Privasea seamlessly integrates cutting-edge technologies to offer a secure, efficient, and collaborative environment.

Design Philosophy

Privasea's architecture is guided by a commitment to:

Privacy-Preserving Machine Learning

The core design principle revolves around Fully Homomorphic Encryption (FHE) technology. FHE allows computations on encrypted data, eliminating the need to expose raw information. This revolutionary approach ensures the privacy and security of user data throughout the entire workflow in model evaluation.

Seamless User Interaction

The Privasea API provides a user-friendly interface, abstracting away the complexities of FHE. This ensures a seamless interaction for users, allowing them to securely submit data, request model training, and obtain predictions.

Decentralized Computation Network

Privanetix - Decentralized Computation Network, comprising high-performance machines with integrated HESea libraries, forms the backbone of Privasea. It provides the computational resources necessary for FHE-based operations on encrypted data. The collaboration among Privanetix nodes enables efficient and scalable execution of privacy-preserving machine learning tasks.

Blockchain-Based Incentives

The Blockchain-based Incentive Module, implemented through smart contracts on the blockchain, plays a pivotal role in fostering collaboration. It tracks registrations, contributions of Privanetix nodes, validates computations, and transparently rewards active participants. This approach ensures contributors are motivated to provide their computational resources while maintaining transparency and fairness throughout the network.

Work Flow

Privasea's workflow ensures a seamless and privacy-focused experience for users engaging with machine learning tasks. The process involves several key steps:

1. User Interaction and Task Initiation:

• Users generate accounts and initiate machine learning tasks.
• Encrypt input vectors locally using the Privacy-Preserving Machine Learning Application API or a DApp integrated with the Privasea Application API.
• Generate a switching key locally.

2. Task Submission to Privanetix:

• Users submit encrypted tasks to Privanetix network.
• Pay service fees securely through blockchain-based transactions.

3. Computation by Privanetix Nodes:

• Privanetix nodes receive and execute encrypted tasks in the user's encryption domain.
• Transfer the encrypted results to the decryptor's encryption domain using the switching key.

4. Result Submission and Reward Withdrawal:

• Privanetix nodes submit the results to decryptors.
• Withdraw service fees for this task through blockchain transactions.

5. Result Decryption by Decryptors:

• Decryptors use their client keys to decrypt the results.

6. Result Delivery to Network User:

• Decrypted results are sent to the network users using Proxy Re-encryption(PRE) scheme.

This workflow ensures end-to-end privacy and security throughout the machine learning task process within the Privasea network.