Skip to content
Crypto.world

Research

EthVault: A Secure and Resource-Conscious FPGA-Based Ethereum Cold Wallet

Commercial cryptocurrency hardware wallets such as Trezor, Ledger, and COLDCARD are physical devices, but their cryptography runs as software on microcontroller units (MCUs) such as the STM32…

EthVault: A Secure and Resource-Conscious FPGA-Based Ethereum Cold Wallet
arXiv
2510.23847
Published
2025-10-27
Authors
Joel Poncha Lemayian, Ghyslain Gagnon, Kaiwen Zhang, Pascal Giard

AI summary

Overview

Research area: Hardware security and applied cryptography for blockchain systems — specifically FPGA-based implementations of Ethereum wallet cryptography.

Technical level: Advanced. The paper assumes familiarity with elliptic curve cryptography, BIP-32/39/44 wallet standards, side-channel analysis, and digital hardware design (LUTs, registers, RAM blocks).

Scope: The paper presents EthVault, a complete FPGA hardware architecture for an Ethereum hierarchical deterministic (HD) cold wallet, including side-channel-resistant elliptic curve cryptography and a hardware child key derivation function, evaluated by resource utilization on a Xilinx Zynq UltraScale+ device.

What This Paper Is About

Commercial cryptocurrency hardware wallets such as Trezor, Ledger, and COLDCARD are physical devices, but their cryptography runs as software on microcontroller units (MCUs) such as the STM32 — leaving them exposed to malware and to physical side-channel attacks that have already extracted private keys in published attacks. The authors argue that moving the sensitive algorithms into dedicated FPGA hardware, configured in a way that is difficult to alter predictably, closes much of that gap.

The goal is a complete, end-to-end hardware design for an Ethereum HD wallet that implements every stage from entropy generation to transaction signing, keeps resource usage low enough for a small portable device, and includes an elliptic curve multiplier whose execution behavior does not vary with the secret key.

Key Contributions

  1. First complete hardware architecture for an Ethereum HD cold wallet. EthVault implements all six stages of the wallet flow — entropy, mnemonic/seed creation, master key derivation, child key derivation, Ethereum address generation, and transaction signing — rather than a single isolated cryptographic primitive.

  2. A SECP256K1 architecture resistant to side-channel and timing attacks. The design uses temporary registers, parallel processing, and branch-free complete elliptic curve point addition (ECPA) formulas to remove the input-dependent variations that differential power analysis (DPA) and timing attacks exploit in a conventional Montgomery ladder.

  3. The first hardware architecture of the child key derivation (CKD) function. CKD is the core of BIP-32 hierarchical key generation and, per the authors, had not previously been implemented in hardware.

  4. A resource-conscious implementation with explicit design targets. The authors set quantifiable goals of under 70k look-up tables (LUTs) and a minimum throughput of 10 kbps (sufficient for real-time signing), derived from prior FPGA implementations of the individual building blocks and from current Ethereum blockchain performance requirements.

Main Findings

  • Low resource footprint on a Xilinx Zynq UltraScale+ FPGA: the complete design uses only 27% of LUTs, 7% of registers, and 6% of RAM blocks.

  • Uniform execution behavior: the proposed ECC architecture exhibits uniform execution behavior across varying inputs, which is the property that underpins its resistance to timing and power-based key recovery.

  • Comparison point for context: the design goal of under 70k LUTs and 10 kbps throughput was set against existing FPGA implementations of the wallet's cryptographic building blocks and against Ethereum's performance requirements; whether the final design meets the 10 kbps target is not stated in the portion of the text available here.

  • Attack surface of current wallets is documented and real: the paper cites private key extraction from a Trezor One wallet via simple power analysis (SPA), successful side-channel attacks on the STM32 MCU used in commercially available wallets, and a modeled adversarial attack that extracted private keys from an isolated wallet in seconds by infecting it with malicious code.

  • Wallet hacks contribute significantly to the billions of dollars lost to cryptocurrency theft, which frames the motivation for hardware-level protection.

  • Algorithm coverage: the architecture includes ECPA with complete addition equations, a Montgomery ladder built from ECPA and ECPD for scalar multiplication, the binary inversion algorithm (BIA) for projective-to-affine conversion, HMAC-SHA-512, CKD, the Ethereum EIP-55 checksum, PBKDF-2, and ECDSA.

Methodology in Plain English

The authors start by decomposing an Ethereum HD wallet into its algorithmic stages: a random number generator produces entropy; BIP-39 turns that into 12 or 24 mnemonic words and, via PBKDF-2 running with 2048 iterations, a 512-bit seed; HMAC-SHA-512 over the seed and the ASCII string "Bitcoin seed" yields the master private key and chain code; the CKD function, using HMAC-SHA-512 and SECP256K1, derives child keys down the BIP-32 tree (Ethereum follows the path m/44'/60'/0'/0/address_index); Keccak-256 produces the address from the public key, keeping the 160 least significant bits, and EIP-55 adds a mixed-case checksum; finally ECDSA signs transactions.

They then build hardware for each stage and deliberately reuse the shared primitives (HMAC-SHA-512 and SECP256K1 appear in several stages) so the total logic stays small. The central security work is in the scalar multiplication: the standard Montgomery ladder is "balanced" in the sense that it performs the same sequence of operations regardless of the key bit, but it writes to different registers depending on whether the bit is 1 or 0, which produces distinguishable power and timing signatures. The authors replace this with temporary registers, parallel processing, and complete addition formulas that eliminate the branching normally needed to handle point doubling and the point at infinity in Weierstrass curve arithmetic.

Finally, they synthesize the design on an FPGA and measure LUT, register, and RAM block utilization, and check that execution behavior is independent of the input values.

Why This Matters

Impact on research: The paper shifts the cryptocurrency wallet discussion from software hardening (secure enclaves like Intel SGX, ARM TrustZone isolation, air-gapping) to dedicated hardware datapaths, and provides what the authors describe as the first hardware CKD implementation — a reusable building block for any BIP-32-compatible hardware wallet, including Bitcoin.

Real-world applications:

  • Portable hardware wallets for consumers, where the low LUT/register/RAM usage supports small form factors.
  • Custody and treasury systems for institutions holding large Ethereum balances, where physical side-channel resistance is a procurement requirement.
  • Hardware signing modules for decentralized finance and smart-contract interaction, where keys must authorize transactions without ever entering a general-purpose computer.
  • Reference designs for FPGA vendors and security-chip makers building blockchain-specific secure elements.

Industry relevance: Existing commercial wallets run on STM32-class microcontrollers with documented side-channel weaknesses. A resource-conscious FPGA architecture that fits comfortably within a mid-range device demonstrates that hardware-level key isolation and constant-behavior ECC are achievable without an impractical silicon budget.

Future Directions

  • Complete and report the performance evaluation. The excerpt ends mid-way through the ECDSA background; throughput, latency, and per-algorithm timing results, plus the direct comparison against the 10 kbps target, are not available in the provided text.

  • Physical validation against real attacks. The paper argues resistance from uniform execution behavior at the architectural level; measured power traces and an attempted DPA/CPA campaign against the implemented design would be the natural confirmation.

  • Address the limitations the authors themselves identify. Section VI of the paper is dedicated to limitations and future work, but its content is not included in the provided excerpt.

  • Extend beyond Ethereum. The architecture already contains the general BIP-32/BIP-39/BIP-44 machinery; supporting additional chains (the paper discusses Bitcoin and SECP256K1 as shared foundations) and other signature schemes is a natural extension.

Target Audience

Hardware security researchers and FPGA/ASIC designers working on cryptographic accelerators; blockchain security engineers evaluating or building custody hardware; cryptography graduate students interested in side-channel-resistant ECC implementation; and technical decision-makers at wallet vendors who need to understand the security argument for moving key operations out of microcontroller software into dedicated logic. Readers without a background in elliptic curve arithmetic or digital design will find Sections II and III demanding.

Authors’ abstract

Cryptocurrency blockchain networks safeguard digital assets using cryptographic keys, with wallets playing a critical role in generating, storing, and managing these keys. Wallets, typically categorized as hot and cold, offer varying degrees of security and convenience. However, they are generally software-based applications running on microcontrollers. Consequently, they are vulnerable to malware and side-channel attacks, allowing perpetrators to extract private keys by…

The full abstract on arXiv
Read the original paper