SPARK

The SPARK project aims to explore and develop the foundation for large-scale, user-operated radio mesh networks. SPARK defines early architecture for decentralized identity, sub-mesh formation, and probabilistic routing under real-world radio and adversarial constraints.

SPARK is in it's earliest stages and is highly experimental. The current implementation focuses on establishing the foundational cryptographic and routing function before optimizing for large scale deployment. At this time, this project is not intended for production use.

Design Goals

• Dynamic infrastructure with no global authority
• Precise and private P2P routing over untrusted intermediate nodes
• Private identities
• Redundant and fallback operations under packet loss, retransmission, and node movement
• Capabilities to transmit data reliably across a nationwide or global infrastructrue built of user-owned routers and clients

Address-based Routing

SPARK uses a cryptographically derived addressing method rather than fixed hardware or positional-based identifiers.

Each node maintains a long-term cryptographic identity. From this identity, nodes derive dynamic, ephemeral addresses that do not reveal the underlying identity. This allows routing decisions to be made without revealing fingerprints or identifiers that could be used for tracking or correlation

Sub-Mesh Design

The network is autonomously and dynamically segmented into sub-meshes. These sub-meshes are based on topology, where reliability and trusted nodes are prioritized utilizing a "trust score"

Sub-mesh boundaries also act as routing domains

• Traffic crossing a boundary incurs additional cryptographic layers, obscuring the identity and route of all data
• Routing confidence decays with distance from the originating sub-mesh
• This allows efficient and private routing without any node acquiring global network knowledge
• Allows for a dynamic and expanding network to thrive without any manual configurations

Packet Structure and Cryptography

SPARK packets use a layered encryption model inspired by extsting onion routing protocols, and adapted for mass deployment across millions of nodes (theoretically)

• Inner payloads are AES-CTR encrypted
• Outer layers utilize AES-GCM for authenticating routing headers
• Routing headers are intentionally mutable between hops and only authenticated at the outer layer
• Each layer utilizes unique nonces to prevent keystream reuse and replay

This structire allows intermediate nodes to make routing decisions without having any knowledge as to the origin, destination, or data contents

To Do:

• Finish porting router firmware to RPi
• Create linking protocol between radios
• Create ESP32 Client firmware
• Create ESP32 client UI (android will be prioritized initially)
• Test basic functionality between a few radios, before simulating network behavior at large scale
• Field test 15-20 routers
• Create basic router UI for diagnostics, configuring, and updates (should NEVER be required for a user to configure or interact with a router directly)