top of page

CUVEX
ARCHITECTURE

Cuvex’s architecture is designed to minimize the attack surface, eliminate external dependencies, and ensure cryptographic sovereignty.

Rather than layering security on top of a traditional model, the system is redesigned from the ground up so that entire classes of attacks are impossible by construction.

ARCHITECTURAL PRINCIPLES

Cuvex’s architecture is based on the following non-negotiable principles:

• Strict separation of responsibilities
• Elimination of backends and centralized services
• Offline signing as the only mode of operation
• Mandatory human verification
• Privacy ensured by technical impossibility, not by promises

SYSTEM COMPONENTS

1. Watch-Only App

Role: transaction monitoring and preparation

• Does not manage private keys
• Does not sign transactions
• It can operate using a self-hosted node or user-selected endpoints via TOR.
• Generates unsigned PSBT transactions

The app is not trusted and it does not need to be.

2. Cuvex Device (Personal HSM)

Role: sovereign encryption and signing

• Processes secrets exclusively in volatile memory
• Does not persist private keys
• Executes all critical operations fully offline
• Requires physical interaction and visual verification

The device has no knowledge of balances, identities, or network context.

3. External Cryptograms (NFC / USB-C)

Role: secret separation

• Seed phrases are stored encrypted (AES-256 GCM)
• The cryptogram never reveals the secret without the device
• The device never persists the secret after use

This eliminates single points of failure.

CHAIN OF TRUST

(SECURE BOOT)

Cuvex implements an explicit root of trust through a secure boot process:

  1. The bootloader resides in a protected region of the device.

  2. Before installing or executing any firmware, the bootloader cryptographically verifies the RSA signature of the binary.

  3. Only authenticated and untampered firmware is allowed to run.

  4. Any unsigned or modified firmware is rejected.

This mechanism prevents:

  • Installation of malicious firmware

  • Downgrade attacks

  • Tampering during manufacturing or distribution

DATA FLOW (HIGH LEVEL OVERVIEW)

  1. The user monitors their wallet in the Watch-Only app.

  2. The app generates an unsigned PSBT transaction.

  3. The PSBT is transferred to the Cuvex device.

  4. The encrypted cryptogram is provided via NFC or USB-C.

  5. The device decrypts the secret in volatile memory and signs the transaction.

  6. The signature is returned without exposing the secret.

At no point:

  • Does the secret touch a network-connected system

  • Does the device access the network

  • Are centralized metadata generated

RELATIONSHIP TO THE THREAT MODEL

Each component exists to mitigate specific threats:

• Watch-Only → malware and compromised applications
• Air gap and PSBT → remote attacks
• External cryptograms → physical attacks
• Secure Boot → supply chain attacks

The architecture is a direct materialization of the threat model.

WHY THIS ARCHITECTURE MATTERS?

Security in self-custody does not depend on exotic algorithms, but on:

  • Isolation

  • Separation

  • Verifiability

Cuvex adopts an architecture-first approach, where cryptography reinforces the design rather than replacing it.

LEARN MORE ABOUT

Cuvex Threat Model

UNDERSTANDING

Privacy by design

COMPARE

Cuvex vs. Hardware Wallets

bottom of page