Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

20 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

AURA — Compact Hardware Root of Trust

182 GE (Base) · ~300 GE SCA-hardened · Silicon-level identity binding · Standard CMOS / FPGA

Patent Simulation Base HRoT SCA-hardened


🌐 Language / 语言 / 言語 / Idioma

English · 简体中文 · 繁體中文 · 日本語 · Español


The Problem

Most embedded devices authenticate in firmware. Whoever controls the software controls the device's identity. For FPGA-based industrial equipment and cost-sensitive IoT endpoints, board-level cloning is a persistent commercial problem, and the available countermeasures are unsatisfying:

  • Bitstream encryption on mainstream FPGA families has published side-channel weaknesses.
  • External authentication chips are cheap but vulnerable to man-in-the-middle interposition on the bus.
  • Full security subsystems (TPM, secure elements, commercial RoT IP) are architected — and priced — for high-assurance silicon. They do not fit the area, power, or licensing budget of the devices where cloning actually happens.

AURA targets that gap: identity binding placed in the logic fabric itself, at a footprint small enough to sit alongside the application.

What AURA Is — and Is Not

AURA is an identity and state anchor, not a cryptographic engine and not a TPM replacement.

It provides: device identity binding, clone detection, a hardware-enforced isolation state that software cannot clear, and SEU anomaly detection.

It does not provide: general-purpose encryption, key storage at secure-element assurance levels, a certified crypto library, or Common Criteria assurance today.

Positioning against adjacent technologies:

Footprint Fits in MCU / small FPGA Primary function
TPM 2.0 ~50,000 GE Full security subsystem
Commercial RoT IP (tRoot, RT-series, PUFrt class) thousands of GE Partial Boot integrity + key management
PUF macro alone ~10,000 GE Partial Unclonable key source
Standalone authentication chip external part Challenge–response auth (bus-exposed)
AURA (Base) 182 GE Identity binding + isolation state
AURA (SCA-hardened) ~280–320 GE Same, with side-channel hardening

This table compares footprint and fit, not equivalent functionality. AURA occupies a different tier: it is not competing with a TPM on assurance, it is addressing designs where a TPM was never an option.

Gate count basis: 182 GE = Base HRoT, measured on FPGA (Vivado 2023.2, Artix-7 35T: 46 LUT + 22 FF). ~280–320 GE = all four SCA layers active (FPGA). ASIC estimate @ 28 nm: ~300–500 GE — estimate, not silicon-measured.

Core Mechanism

Ternary-state encoding over standard binary CMOS — no special process required.

State Meaning Behaviour
2'b01 Legitimate Normal operation
2'b10 Isolated Access blocked; authorized owner can recover via credential verification
2'b11 Illegal / Alert Anomaly or SEU event detected

Key property: once 2'b10 enters the MAC chain, no software instruction can clear it. Recovery requires credential verification through a hardware-defined path.

This is deliberate, and it is not a brick. For the product owner, an isolation event is a serviceable condition with a defined recovery procedure — and a recoverable one, which means it can be operated as a support workflow rather than an RMA. For an attacker holding a cloned unit, there is no software path back.

Four-Layer Side-Channel Hardening

Layer Mechanism Target
L1 Masked LFSR Differential power analysis
L2 Dual-rail logic DPA, fault injection, SEU (0-cycle detection)
L3 Constant-time MAC Timing side channels
L4 Random delay insertion DPA trace alignment

All four layers operate in RTL — no firmware dependency.

Verification status. These mechanisms are implemented and functionally verified in simulation. Physical side-channel evaluation (TVLA, per ISO/IEC 17825) is in progress. Until that data is published here, no claim of measured attack resistance should be considered validated. We would rather state this plainly than have an evaluator discover it.

By the Numbers

Metric Value Basis
Gate count — Base HRoT (FPGA) 182 GE (46 LUT + 22 FF) Measured, Vivado 2023.2, Artix-7 35T
Gate count — SCA-hardened (FPGA) ~280–320 GE Measured
ASIC @ 28 nm ~300–500 GE Estimate
Silicon area @ 28 nm < 0.003 mm² Estimate
Power < 1 mW Estimate
Functional simulation 22/22 scenarios PASS Icarus Verilog
RTL modules 9 complete
Physical SCA (TVLA) In progress

Validation Artifacts

Architecture

  • System overview — SCA Defense Layer → Aura 2 (sense) → ESM (decide) → Aura 1 (execute)
  • Attack type → defense mechanism → security effect mapping
  • Power-on boot check and isolation-state write flow — state survives power cycling
  • Single-rail vs. dual-rail power profile comparison

See docs/.

Hardware Demonstration

Live recordings on Artix-7 35T covering normal authentication, repeated-failure isolation, power-cycle persistence, authorized recovery, replay rejection, and privilege-escalation blocking. Recorded on real hardware with UART output at 115200 baud — not simulation.

Publishing to GitHub Releases. ⭐ Star to be notified.

Who This Is For

FPGA-based equipment manufacturers — industrial controllers, instrumentation, medical and test equipment facing board-level cloning. Deployable on existing hardware today; no respin required. This is our nearest-term focus.

IoT and embedded silicon — smart locks and high-security endpoints where a full security subsystem does not fit the area or power budget. EU EN 18031 became applicable under the Radio Equipment Directive in August 2025; SESIP L2 is a supported certification path.

Automotive — 4-layer SCA hardening addresses ISO 21434 hardware prerequisites, at a small area penalty on ECU silicon. Longer qualification cycle; engaged as a strategic path rather than a near-term one.

RISC-V SoC projects — AMBA APB integration wrapper under development.

Getting Started

FPGA evaluation (Artix-7 / Basys 3) — contact us → NDA → evaluation package: RTL interface definitions, integration docs, simulation scripts. Typical evaluation: 2–4 weeks.

ASIC integration — engage 3–6 months before tape-out. Full RTL package, synthesis constraints, and timing reports provided under NDA.

Intellectual Property

  • Chinese invention patent application No. 202610850983.0 (filed 2026)
  • Chinese invention patent application No. 2026106956971 (filed May 2026)
  • PCT filing in progress — CN / US / EU / JP / KR

Source in this repository is published for technical evaluation only. Commercial use requires a separate written license. See LICENSE.

Contact

OptiAura Tech — 上海若爻高科技有限公司

📩 lexxu@optiaura.tech · 🌐 optiaura.tech · 👤 Lex Xu on LinkedIn

Full RTL review available under NDA. Integration engineering support provided throughout evaluation.

About

Compact hardware root of trust — 182 GE silicon-level identity binding for FPGA and IoT

Topics

Resources

Stars

1 star

Watchers

1 watching

Forks

Releases

Packages

Contributors