Architecture Overview

The Arcium Network is engineered for secure, distributed confidential computing across a wide range of applications, from AI to decentralized finance (DeFi) and beyond. It is built on advanced cryptographic techniques, including Multi-Party Computation (MPC), to enable trustless and verifiable computations without the need for a central authority.

1. Multi-Party Execution Environments (MXEs)

MXEs are dedicated, compartmentalized environments where computations are defined and securely executed. They allow for parallel processing (as Clusters can concurrently compute for various MXEs), improving throughput and security.

MXEs are highly configurable, enabling Computation Customers to define their security requirements, encryption schemes, and performance parameters. While individual computations are executed within specific Clusters of Arx Nodes, multiple Clusters can be associated with a single MXE.

This ensures that computations can be reliably executed, even if certain Nodes within a Cluster are offline or experiencing high load. By predefining these configurations, customers can tailor the environment to meet their specific use cases with a high degree of flexibility.

1. arxOS

arxOS is the distributed execution engine that coordinates computations within the Arcium Network, powering Arx Nodes and Clusters. Each node (like a core in a computer) provides computational resources to execute computations defined by MXEs.

1. Arcis (Arcium's Developer Framework)

Arcis (a Rust-based developer framework) enables developers to build on top of the infrastructure, supporting all of Arcium's MPC protocols. It consists of a Rust-based framework and compiler.

1. Clusters of Arx Nodes (running arxOS)

arxOS is the distributed execution engine that coordinates computations within the Arcium Network, each node (like a core in a computer) provides computational resources to execute computations defined by MXEs.

Clusters offer customizable trust models, supporting both dishonest majority protocols (initially Cerberus) and "honest but curious" protocols (see Manticore). Other protocols, including honest majority protocols, will be added in the future for additional use case scenarios.

Chain-Level Enforcement

All state management and orchestration of computations are handled on-chain via Solana, serving as a consensus layer that coordinates the actions of Arx Nodes. This ensures fair reward distribution, enforcement of network rules, and alignment among Nodes on the current state of the network. Tasks are queued within a decentralized mempool architecture, where the on-chain component helps determine which computations have the highest priority, identify misbehavior, and manage the order of execution. Read more on the on-chain architecture here.

Nodes stake collateral to ensure compliance with network rules. In the case of misbehavior or deviations from protocol, slashing is implemented to penalize and maintain the integrity of the network. See the Staking Overview section for more staking and slashing.