Introducing the Unified BCI Device Explorer

The Problem

If you want to research BCI security today, you have to piece it together yourself. Device specs are scattered across FDA 510(k) filings, manufacturer marketing pages, and paywalled journal papers. Threat models for neural interfaces don’t exist in any structured, machine-readable form. The physics constraints that govern what BCIs can and cannot do are buried across thermodynamics textbooks, electromagnetics references, and neuroscience papers that rarely cite each other.

There is no single place where a researcher, developer, or policymaker can look up a BCI device, see what brain regions it targets, understand what attacks apply to it, check which physics constraints limit its design, and pull all of that into their own analysis.

We built that place.

What We Built

The BCI Research Hub is now live at qinnovate.com/bci. It consolidates everything the QIF (Quantum Information Framework) project has catalogued about brain-computer interface hardware, physics constraints, and security into a single, searchable, cross-referenced research tool.

This is not a product page. Every data point links back to its source: FDA filings, manufacturer datasheets, peer-reviewed papers, or the QIF derivation log where it was first documented. Every number has a confidence level. Every cross-reference is explicit.

The hub has four sections, each addressing a different layer of BCI research. They all feed into a single open API.

The Four Sections

1. Device Explorer

URL: qinnovate.com/bci/explorer

24 BCI devices catalogued from FDA filings, manufacturer datasheets, and peer-reviewed literature. Each device card includes:

• Full hardware specs: electrode type, channel count, sampling rate, data rate, signal bandwidth
• Modality: EEG, ECoG, intracortical, fNIRS, TMS, focused ultrasound
• Invasiveness tier: non-invasive, minimally invasive, invasive
• FDA status: cleared, breakthrough designation, investigational, research-only
• Target brain regions: which of the 38 mapped brain structures each device can reach
• Mapped threat techniques: which of the 103 TARA techniques apply to this device’s QIF band
• Confidence level: how well-sourced each data point is (peer-reviewed, manufacturer-stated, inferred)

The explorer is searchable and filterable. You can find all non-invasive EEG devices with 32+ channels, or all invasive devices with FDA breakthrough designation, or all devices that operate in the N1 (electrode-tissue) band of the QIF model.

Devices span the full range: consumer headsets like the Emotiv EPOC X and Muse 2, research-grade systems like the g.tec g.USBamp and ANT Neuro eego, clinical devices like the NeuroPace RNS and Medtronic Percept PC, and cutting-edge implants like Neuralink N1, Synchron Stentrode, and Blackrock MicroPort’s Utah Array.

2. BCI Limits Equation

URL: qinnovate.com/bci/limits

Before you can secure a BCI, you need to know what physics allows. We synthesized 13 constraints from thermodynamics, electromagnetism, information theory, Moore’s Law scaling, and biocompatibility into a unified system that defines the fundamental design envelope for any brain-computer interface.

The 13 constraints are grouped into five categories:

Thermodynamics & Power:

• Landauer limit (minimum energy per bit erasure)
• Thermal ceiling (2C max brain temperature rise, Penne bioheat equation)
• On-chip vs. telemetry power trade-off (40 mW implant budget)
• Margolus-Levitin bound (maximum computational speed per unit energy)

Electromagnetic & Wireless:

• EM propagation (tissue attenuation at target frequencies)
• Wireless bandwidth (Shannon capacity under neural-safe power limits)

Scaling & Geometry:

• Moore’s Law scaling (electrode density doubling time ~7.4 years)
• Geometric fit (cortical curvature vs. rigid electrode arrays)
• Information-theoretic ceiling (simultaneous neuron recording limits)

Safety & Biocompatibility:

• Shannon electrode safety (charge injection limits per electrode area)
• Mechanical mismatch (brain tissue ~1.5 kPa vs. silicon ~170 GPa)

Signal & Detection:

• Boltzmann detectability (minimum detectable signal vs. Johnson-Nyquist noise at 310K)
• QIF coherence floor (minimum signal integrity for meaningful BCI operation)

Each constraint page shows the equation, a plain-English explanation, the physical constant values used, and how it connects to BCI design decisions.

No published paper unifies all 13 of these constraints into a single system. This is original work, documented in Entry 60 of the QIF Derivation Log. The system was independently cross-validated by Gemini 2.5 Pro, which verified 12 of 13 constraints as correct and identified 2 corrections (mechanical mismatch ratio inverted, Johnson noise temperature corrected to 310K). Both corrections were applied.

3. Security Guardrails

URL: qinnovate.com/bci/guardrails

The on-device neurosecurity stack that QIF proposes for BCI signal integrity. Four layers:

• L0: Impedance Guard - Hardware-level electrode impedance monitoring. Detects physical tampering, electrode degradation, and injection attacks at the analog front-end.
• L1: Coherence Monitor - The core of Neurowall. Computes a real-time coherence score (Cs) from phase variance, spectral entropy, and cross-channel correlation. Detects signal injection, spectral manipulation, and feedback loop attacks.
• L2: Thermal Budget - Enforces the 2C thermal ceiling from the BCI Limits Equation. Prevents sustained high-power attacks that could cause tissue damage.
• L3: Protocol Envelope - NSP (Neural Security Protocol) session validation. Post-quantum key exchange (ML-KEM-768) and signatures (ML-DSA-65) for authenticated neural data transport.

The guardrails page also includes a “What Exists vs. What’s New” breakdown showing which defenses are established practice and which are original QIF contributions, along with a six-phase implementation roadmap.

4. API Documentation

URL: qinnovate.com/bci/api

Full documentation for the unified QIF API, including endpoint details, response structure, usage examples in JavaScript, Python, and curl, and information about legacy endpoints.

The Unified QIF API

Everything in the BCI Research Hub, plus the full TARA threat registry, brain atlas, scoring specifications, and project timeline, is available through a single open endpoint. No authentication. No rate limits. No API key. CORS enabled for browser-side use.

Endpoint

GET https://qinnovate.com/api/qif.json

Full API documentation is available with interactive examples.

What the API Returns

Section	Description	Count
hourglass_bands	QIF 11-band model (N7-N1, I0, S1-S3)	11
threats.techniques	TARA attack/therapeutic techniques	103
threats.categories	Threat categories	13
threats.tactics	MITRE-compatible tactics	13
devices.inventory	BCI hardware with full specs	24
brain_atlas.regions	Mapped brain structures	38
brain_atlas.device_mappings	Device-to-region cross-references	per device
brain_atlas.neural_latency	Neural pathway timing metrics	per region
physics.constraints	BCI Limits Equation constraints	13
physics.constants	Verified physical constants	14
specs.niss	Neural Impact Scoring System specification	1
specs.tara	TARA registry specification	1
specs.dsm5	DSM-5-TR diagnostic mapping specification	1
timeline	Project milestones and discoveries	31
current_stats	Live dataset statistics	1

Total payload: ~580 KB. Cached for 1 hour.

Quick Start

JavaScript:

const qif = await fetch('https://qinnovate.com/api/qif.json').then(r => r.json());

// All devices with their brain region mappings
qif.devices.inventory.forEach(d => {
  console.log(`${d.name}: ${d.channels}ch ${d.modality}, regions: ${d.regions?.map(r => r.name).join(', ')}`);
});

// Cross-reference: which threats apply to intracortical devices?
const invasiveDevices = qif.devices.inventory.filter(d => d.invasiveness === 'invasive');
const invasiveBands = [...new Set(invasiveDevices.flatMap(d => d.qif_bands))];
const relevantThreats = qif.threats.techniques.filter(t => invasiveBands.includes(t.band));

Python:

import requests

qif = requests.get('https://qinnovate.com/api/qif.json').json()

# Physics constraints grouped by category
for c in qif['physics']['constraints']:
    print(f"[{c['category']}] {c['name']}: {c['equation']}")

# Devices by FDA status
from collections import Counter
fda = Counter(d.get('fda_status', 'unknown') for d in qif['devices']['inventory'])
print(fda)

# High-severity threats with DSM-5 mappings
for t in qif['threats']['techniques']:
    if t.get('severity') == 'critical':
        print(f"{t['id']}: {t['name']} — DSM-5: {t.get('dsm5_codes', 'unmapped')}")

curl + jq:

# Device summary table
curl -s https://qinnovate.com/api/qif.json | \
  jq -r '.devices.inventory[] | [.name, .modality, .channels, .fda_status] | @tsv'

# All physics constraint equations
curl -s https://qinnovate.com/api/qif.json | \
  jq '.physics.constraints[] | {name, equation, category}'

# Project timeline
curl -s https://qinnovate.com/api/qif.json | \
  jq '.timeline[] | "\(.date) [\(.type)] \(.title)"'

How It All Connects

The key design decision behind the BCI Research Hub is cross-referencing. Every dataset shares the same identifiers:

• QIF band IDs (N7, N6, N5, N4, N3, N2, N1, I0, S1, S2, S3) link devices to threats to physics constraints. A device operating at band N1 (electrode-tissue interface) faces a specific subset of the 103 TARA techniques, is governed by specific physics constraints (Shannon electrode safety, mechanical mismatch, impedance), and maps to specific brain regions.

• TARA technique IDs (QIF-T0001 through QIF-T0103) link threats to devices, brain regions, DSM-5-TR diagnostic codes, neurorights implications, FDORA regulatory coverage, and physics feasibility tiers. Each technique page is at qinnovate.com/TARA/QIF-T{id}.

• Brain region mappings link the 38 structures in the Brain-BCI Atlas to specific devices, neural latency metrics, and QIF bands.

This cross-referencing is what makes the dataset useful beyond a simple catalog. You can start from any entry point, a device, a threat, a brain region, a physics constraint, and trace connections across the entire system.

Related Work

The BCI Research Hub builds on several other QIF components that are also publicly available:

• TARA Registry - The full 103-technique threat and therapeutic atlas. Each technique has its own page with NISS scoring, DSM-5-TR mappings, neurorights analysis, physics feasibility tier, and FDORA regulatory coverage. The TARA blog post explains the dual-use reframing.

• Brain-BCI Atlas - Interactive 3D visualization of 38 brain structures mapped to BCI devices and QIF bands.

• QIF Security Model - The 11-band hourglass model, coherence scoring pipeline, and the four-step Map-Score-Protect-Deliver framework.

• Neural Security Protocol (NSP) - Post-quantum secure transport protocol for neural data. ML-KEM-768 key exchange, ML-DSA-65 signatures, 5-layer neurosecurity stack. Rust implementation available.

• Neurowall - Signal integrity monitor simulation. 7 of 9 attack types detected at 15 seconds, 9 of 9 at 30 seconds. ROC-optimized operating point: 5% false positive rate, 100% true positive rate at threshold=12, duration=20s.

• NISS Scoring - The Neural Impact Scoring System that quantifies threat severity across reversibility, detectability, consent violation, and clinical impact dimensions.

• DSM-5-TR Psychiatric Mapping - The Neural Impact Chain that bridges TARA attack techniques to DSM-5-TR diagnostic codes, showing how neural attacks map to potential psychiatric outcomes.

• Zenodo Preprint - The academic paper (v1.4, 28 pages, 6 figures, CC-BY 4.0) documenting the full QIF framework.

• Validation Dashboard - Live tracking of cross-AI validation sessions, citation verification, and fact-checking status.

Why Open

BCI security is too important to be locked behind paywalls or proprietary databases. The devices listed in this explorer are being implanted in human brains right now. The attack techniques in TARA are not theoretical, they are grounded in published research with demonstrated feasibility. The physics constraints are laws of nature, not trade secrets.

We compiled it, verified it, cross-referenced it, and put it behind a single GET request with no authentication. The source data is on GitHub. The API is at qinnovate.com/api/qif.json. The API documentation explains every field.

Use it. Build on it. If you find an error, open an issue. If you want to add a device or technique, submit a pull request.

What’s Next

• More devices as new FDA filings and peer-reviewed papers come out
• Real-time data feeds from Neurowall simulation runs
• STIX/TAXII format export for integration with existing cybersecurity tooling (already available at /api/stix.json)
• Community contributions via pull request
• Interactive cross-reference explorer (click a device, see its threats, brain regions, and physics constraints in one view)

A Personal Note

Earlier this year, I made a commitment to myself: get sober, get serious, make my dreams come true. I drew a mind map. Mapped out my future, what I care about, what gets me excited. Ironically, what I landed on was mapping the mind.

I started with 60 attack techniques collectively known across BCI research publications, the baseline of what the field had already identified in scattered papers and conference proceedings. The techniques I added after that came from applying the QIF framework I devised, grounding new attack vectors in physics and neuroanatomy rather than cataloguing what others had already published. All the while learning neuroscience and neurorights from scratch. Here’s where things stand:

• 103 threat techniques mapped to 68 DSM-5 psychiatric diagnoses and 4 neurorights
• 24 BCI devices catalogued from FDA filings and manufacturer datasheets
• 38 brain regions cross-referenced to devices, threats, and physics constraints
• 13 physics constraints unified into a single BCI limits equation, cross-validated by Gemini 2.5 Pro
• An 11-band hourglass security model derived from scratch
• A post-quantum neural security protocol (NSP v0.5) with ML-KEM-768 handshake and ML-DSA-65 signatures implemented in Rust (cryptographic core complete, frame pipeline in progress)
• A coherence monitor (Neurowall v0.7) with 14 attack generators, 9/9 detection at 30s, validated via BrainFlow synthetic board (16-channel, 0% false positive rate), ROC-optimized to 5% FPR at 100% TPR
• A domain-specific neural rendering language (Runemate)
• FDORA Section 3305 regulatory compliance mapped across all techniques
• 9 cross-AI validation sessions across Claude, Gemini, and OpenAI
• A peer-reviewed preprint on Zenodo (v1.4, DOI: 10.5281/zenodo.18640105)
• 165 research sources, 70 derivation log entries, 20 field journal entries, 50 blog posts
• A unified open API serving all of it at a single endpoint

People might ask why I laid all of this groundwork before getting into the actual BCI security work. David Marr answered that in 1982: “Trying to understand perception by studying only neurons is like trying to understand bird flight by studying only feathers; it just cannot be done.” (via MIT OCW 9.13, Nancy Kanwisher). That is the reason all of this exists. You cannot secure a brain-computer interface by looking at the interface alone. You need the psychiatry to understand what damage looks like. You need the physics to understand what’s possible. You need the ethics to understand what’s worth protecting. And you need security to stress test and map it all. There’s more that I’m missing, and I’ll need to disseminate it clearly as I work to improve and enhance these datasets, make them more effective and testable, solidified with research and science. But the foundation had to come first.

There’s a lot of policy gaps that still need addressing. Policies define the scope and the work for security. Neuroethics will write those policies, and neurosecurity will test and secure against them. The two aren’t separate disciplines, they’re two halves of the same problem. Without ethics defining what’s protected, security has no boundaries. Without security testing those boundaries, ethics stays theoretical.

I haven’t kicked off the security vulnerability scanning yet. That’s a whole new category: responsible disclosures, compiled reports, coordinated follow-ups. I can automate parts of it, but that defeats the fun of learning the process.

This isn’t a solo mission forever. I want to build a real team around neurosecurity, but through academia, with proper footing and a real foundation to grow from. That’s my goal by end of year: get into a program and start these conversations in a formal setting.

I started this project with ethics, not exploits. That’s what led to everything you see here. All of this is open for a reason. Don’t wait to start, because I didn’t. Just do it responsibly, securely, and ethically. Make mistakes, but know how to steer yourself back. Version control is beautiful for that reason. Production environments in a lab, that’s different. That requires partnerships, institutional backing, and real accountability.

What a time to be alive. It started with a mind map. I hope it helps science, research, and a whole lot more.

---

Written with AI assistance (Claude). All claims verified by the author.