oni-framework 0.2.0

Neural security framework for brain-computer interfaces. Calculate signal trust scores, filter suspicious signals, anonymize neural data, and validate BCI communications against the 14-layer ONI model.

ONI Framework

Organic Neural Interface Framework - A Python library for brain-computer interface security.

Research Status: This library implements theoretical frameworks for BCI security that have not yet been empirically validated. It is intended for research, experimentation, and to provide shared vocabulary for the emerging field of neural interface security. The mathematical models (Cₛ coherence metric, f × S ≈ k invariant) are derived from neuroscience literature but require experimental validation against real BCI data. Contributions and validation efforts are welcome.

---

Important: What This Library Is (and Isn't)

This is a framework, not a measurement tool.

The ONI Framework provides mathematical formulas and models for evaluating neural signal trustworthiness. It does NOT:

• Connect to any hardware
• Measure signals from your computer or any device
• Collect real data

All examples use sample data. When you see code like:

arrival_times = [0.0, 0.025, 0.050, 0.075]
amplitudes = [100, 98, 102, 99]

These are made-up numbers for demonstration. In a real application, this data would come from:

• A BCI device (like Neuralink, OpenBCI, etc.)
• Neural signal processing software
• Research lab equipment

Some values are theoretical defaults. The transport reliability factors (synaptic transmission = 0.85, etc.) are based on neuroscience literature, not live measurements. In a real BCI system, you would measure these from your specific hardware and tissue interface.

Think of this library as:

• A calculator that can process neural signal data (once you have it)
• A shared vocabulary for discussing BCI security concepts
• A starting point for building real security systems

---

Overview

The ONI Framework provides tools for validating and securing neural signals at the brain-computer interface boundary. It implements:

• Coherence Metric (Cₛ) - Quantify signal trustworthiness across timing, pathway, and amplitude dimensions
• 14-Layer Model - Unified architecture bridging biological (L1-L7) and silicon (L9-L14) domains
• Neural Firewall - Zero-trust signal filtering at the Neural Gateway (L8)
• Scale-Frequency Invariant - Validate signals against the f × S ≈ k constraint

Installation

# From PyPI (recommended)
pip install oni-framework

# With visualization support
pip install oni-framework[viz]

# From source (for development)
git clone https://github.com/qikevinl/ONI.git
cd ONI/MAIN/oni-framework
pip install -e ".[dev]"

Quick Start

Calculate Coherence Score

What it does: Calculates a "trust score" (0 to 1) for a neural signal based on timing consistency, pathway reliability, and amplitude stability.

What the data means:

• arrival_times — When each signal pulse arrived (in seconds). In a real BCI, this would come from your device's timestamp data.
• amplitudes — How strong each pulse was (in microvolts). In a real BCI, this would be the measured voltage at each electrode.
• reference_freq — The expected brain oscillation frequency. 40 Hz (gamma waves) is used for cognitive processing signals.

from oni import CoherenceMetric, calculate_cs

# Create metric with 40 Hz gamma reference
metric = CoherenceMetric(reference_freq=40.0)

# SAMPLE DATA (not real measurements!)
# In a real application, these would come from your BCI device
arrival_times = [0.0, 0.025, 0.050, 0.075, 0.100]  # seconds
amplitudes = [100, 98, 102, 99, 101]  # μV

# Calculate coherence
cs = metric.calculate(arrival_times, amplitudes)
print(f"Coherence Score: {cs:.3f}")

# Interpret the score
level, description = metric.interpret(cs)
print(f"{level}: {description}")

What the score means:

• 0.8 - 1.0: High coherence — signal is consistent and trustworthy
• 0.5 - 0.8: Medium coherence — some variance, may need verification
• 0.0 - 0.5: Low coherence — signal is inconsistent, possibly tampered or noisy

Use the Neural Firewall

What it does: Acts like a security guard for neural signals. It evaluates each signal and decides whether to ACCEPT, FLAG, or REJECT it based on coherence score and authentication status.

Real-world analogy: Like a firewall on your computer that blocks suspicious network traffic, this would block suspicious neural signals before they reach the brain or the computer interpreting brain signals.

What the parameters mean:

• threshold_high (0.6) — Signals above this coherence score are considered trustworthy
• threshold_low (0.3) — Signals below this are automatically rejected
• amplitude_bounds — Acceptable voltage range; anything outside is rejected (prevents over-powered attacks)
• authenticated — Whether the signal source has been verified (like a password check for the device)

from oni import NeuralFirewall
from oni.firewall import Signal

# Create firewall with default thresholds
firewall = NeuralFirewall(
    threshold_high=0.6,
    threshold_low=0.3,
    amplitude_bounds=(0, 500),  # μV limits
)

# SAMPLE DATA (not real measurements!)
# In a real application, this would come from your BCI device
signal = Signal(
    arrival_times=[0.0, 0.025, 0.050],
    amplitudes=[100, 98, 102],
    authenticated=True,  # Device identity verified
)

# Filter the signal
result = firewall.filter(signal)

print(f"Decision: {result.decision.name}")  # ACCEPT, ACCEPT_FLAG, or REJECT
print(f"Coherence: {result.coherence:.3f}")
print(f"Alert Level: {result.alert_level.name}")
print(f"Reason: {result.reason}")

Possible decisions:

• ACCEPT — Signal is trusted, allow it through
• ACCEPT_FLAG — Signal is borderline, allow but log for review
• REJECT — Signal is untrusted or suspicious, block it

Explore the 14-Layer Model

What it does: Provides a conceptual map of how information flows between the brain and a computer. Think of it like a building with 14 floors — signals travel up and down through each layer.

Why 14 layers? The traditional OSI network model has 7 layers. ONI extends this with 7 biological layers (brain side) + 1 bridge layer (where brain meets device) + 6 silicon layers (computer side) = 14 total.

This is a reference model, not code that processes signals. It helps researchers and engineers speak the same language when discussing where attacks might happen or where defenses should be placed.

from oni import ONIStack

stack = ONIStack()

# Print the stack diagram
print(stack.ascii_diagram())

# Access specific layers
gateway = stack.layer(8)  # Neural Gateway — where brain meets device
print(f"Layer 8: {gateway.name}")
print(f"Function: {gateway.function}")
print(f"Attack surfaces: {gateway.attack_surfaces}")

# Iterate through biological layers (L1-L7, brain side)
for layer in stack.biological_layers():
    print(f"L{layer.number}: {layer.name}")

Layer summary:

• L1-L7 (Silicon/OSI): Standard networking layers — the computer's data movement
• L8 (Neural Gateway): The critical boundary where electrodes meet neurons — this is where the firewall operates
• L9-L14 (Biology): From signal processing to identity — the brain's neural processing

Validate Scale-Frequency Relationship

What it does: Checks if a signal's frequency makes sense for its spatial scale. The brain follows a pattern: small structures (neurons) oscillate fast, large structures (brain regions) oscillate slowly.

Real-world analogy: A hummingbird's wings beat fast (small), an elephant's legs move slowly (large). If you saw an elephant moving its legs 100 times per second, you'd know something was wrong. This module catches similar "impossible" neural signals.

The formula: frequency × spatial_scale ≈ constant (k)

Why this matters for security: An attacker injecting fake signals might use the wrong frequency for the brain region they're targeting. This check catches that mismatch.

from oni import ScaleFrequencyInvariant

sfi = ScaleFrequencyInvariant()

# SAMPLE CHECK (not a real measurement!)
# Ask: "Does a 40 Hz signal at 100 μm scale make biological sense?"
frequency = 40  # Hz (gamma wave)
spatial_scale = 1e-4  # meters (100 μm = size of a small neural cluster)

is_valid = sfi.validate(frequency, spatial_scale)
deviation = sfi.deviation(frequency, spatial_scale)

print(f"Valid: {is_valid}")
print(f"Deviation from expected: {deviation:.1%}")

# What frequency SHOULD we see at a given scale?
expected_f = sfi.expected_frequency(spatial_scale=1e-3)  # 1mm
print(f"Expected frequency at 1mm scale: {expected_f:.1f} Hz")

# Print the full hierarchy of scales and expected frequencies
print(sfi.hierarchy_report())

What the hierarchy shows:

Scale	Size	Expected Frequency	Example
Molecular	~10 nm	Very fast	Ion channels
Cellular	~10 μm	~100-1000 Hz	Single neurons
Regional	~10 mm	~1-10 Hz	Brain regions
Whole-Brain	~100 mm	<1 Hz	Global states

Core Concepts

Coherence Metric Formula

Cₛ = e^(−(σ²φ + σ²τ + σ²γ))

In plain English: The coherence score is calculated from three types of "variance" (inconsistency). More variance = lower score = less trustworthy.

Component	Symbol	What It Measures	Data Source
Phase variance	σ²φ	Timing jitter — are pulses arriving when expected?	Measured from signal — compares arrival times to reference frequency
Transport variance	σ²τ	Pathway reliability — how lossy is the signal path?	Model defaults — based on neuroscience literature (can be overridden)
Gain variance	σ²γ	Amplitude stability — is signal strength consistent?	Measured from signal — compares each amplitude to the mean

Important: Transport variance uses hardcoded default values from neuroscience research:

DEFAULT_TRANSPORT_FACTORS = {
    'myelinated_axon': 0.999,       # Very reliable (insulated nerve fibers)
    'unmyelinated_axon': 0.97,      # Slightly less reliable
    'synaptic_transmission': 0.85,  # Synapses sometimes fail to fire
    'dendritic_integration': 0.90,  # Some signal loss in dendrites
}

In a real BCI system, you would measure these from your specific hardware and replace the defaults.

Firewall Decision Matrix

Coherence	Authentication	Decision
High (>0.6)	Valid	ACCEPT
High (>0.6)	Invalid	REJECT
Medium (0.3-0.6)	Valid	ACCEPT + FLAG
Medium (0.3-0.6)	Invalid	REJECT
Low (<0.3)	Any	REJECT

Scale-Frequency Invariant

f × S ≈ k (constant)

Higher frequencies → Smaller spatial scales
Lower frequencies → Larger spatial scales

Project Structure

oni-framework/
├── oni/
│   ├── __init__.py      # Package exports
│   ├── coherence.py     # Cₛ calculation
│   ├── layers.py        # 14-layer model
│   ├── firewall.py      # Signal filtering
│   └── scale_freq.py    # f × S ≈ k invariant
├── tests/
│   └── test_*.py        # Unit tests
├── pyproject.toml       # Package configuration
└── README.md            # This file

API Reference

oni.CoherenceMetric

• calculate(arrival_times, amplitudes) → Coherence score (0-1)
• calculate_variances(...) → Individual variance components
• interpret(cs) → (level, description) tuple

oni.NeuralFirewall

• filter(signal) → FilterResult with decision
• filter_batch(signals) → List of FilterResults
• get_stats() → Filtering statistics
• register_callback(level, fn) → Alert callbacks

oni.ONIStack

• layer(n) → Get layer by number (1-14)
• biological_layers() → L1-L7
• silicon_layers() → L9-L14
• bridge_layer() → L8 (Neural Gateway)
• ascii_diagram() → Visual representation

oni.ScaleFrequencyInvariant

• validate(frequency, scale) → Boolean validity
• deviation(frequency, scale) → Fractional deviation
• expected_frequency(scale) → Predicted frequency
• anomaly_score(frequency, scale) → 0-1 anomaly score

Research Background

This library implements concepts from the ONI Framework research:

• ONI Framework Overview
• Coherence Metric Technical Document
• Neural Firewall Architecture
• Scale-Frequency Invariant

Contributing

Contributions welcome! See CONTRIBUTING.md for guidelines.

Seeking input from:

• Neuroscientists — Validate biological assumptions
• Security engineers — Identify attack vectors
• BCI researchers — Test against real data

License

Apache License 2.0 - See LICENSE

Citation

If you use this library in research, please cite:

@software{oni_framework,
  author = {Qi, Kevin L.},
  title = {ONI Framework: Security Library for Brain-Computer Interfaces},
  year = {2026},
  url = {https://github.com/qikevinl/ONI}
}