Skip to content

murtazaai/polar-bear-biochip

Repository files navigation

polar-bear-biochip

Bio-chip intelligence framework for Polar Bear Systems
Multi-sensor EEG + motion fusion · rig-core LLM cognitive inference · ECDSA-signed provenance

Rust Edition rig-core License: PBS

Built by Murtaza Ali Imtiaz · Technology Lead · Polar Bear Systems · July 2019–Present


Overview

Core component of the Polar Bear Systems superpower bio-chip intelligence platform: bridges neurotechnology with decentralised AI infrastructure.

Fuses real-time EEG brainwave data (δ θ α β γ) with 3-axis MEMS accelerometer readings into higher-order cognitive features, then routes them through a rig-core (ARC) LLM agent for classification. Every inference result is ECDSA-signed on secp256k1 for blockchain-grade, tamper-evident data provenance.

Demo mode: all operations run without an API key by default.
Set ANTHROPIC_API_KEY (and optionally --features ai-agent) for live inference.


Architecture

┌─────────────────────────────────────────────────────────────────────┐
│                      polar-bear-biochip                             │
├──────────────────────────────────┬──────────────────────────────────┤
│         sensors layer            │        provenance layer          │
│                                  │                                   │
│  bci.rs       δ θ α β γ bands   │  ecdsa_signer.rs  secp256k1     │
│               attention + med    │  EcdsaSigner      sign_result()  │
│  accelerometer.rs  3-axis MEMS  │  EcdsaVerifier    from_hex()     │
│               activity state     │  SHA-256 payload hashing        │
│  fusion.rs    SensorFusion       │  64-byte r‖s compact signature  │
│               cognitive_load     │  → SignedOutput JSON on disk     │
│               emotional_valence  │                                   │
│               arousal_level      │                                   │
├──────────────────────────────────┴──────────────────────────────────┤
│         agent layer  (feature = ai-agent)                           │
│                                                                      │
│  biochip_agent.rs   BioChipAgent                                    │
│  ├── ai-agent feature ──► rig-core 0.37 · claude-sonnet-4-6        │
│  └── fallback        ──► curl subprocess · same JSON payload        │
└─────────────────────────────────────────────────────────────────────┘

EEG Signal Processing

Frequency bands (Berger / Niedermeyer taxonomy)

Band Range (Hz) Cognitive correlate
Delta 0.5–4 Deep sleep, unconscious processing
Theta 4–8 Drowsiness, creativity, memory encoding
Alpha 8–12 Relaxed alertness, idle visual cortex, flow states
Beta 12–30 Active thinking, focus, problem-solving
Gamma 30–100 High-level cognition, cross-cortical binding

Derived cognitive features

Feature Formula Range
Attention index β / (α + θ + ε) × 0.6 [0, 1]
Meditation index α / (β + γ + ε) × 4.0 [0, 1]
Cognitive load β / (α + θ + ε) × 0.5 + activity_boost [0, 1]
Emotional valence (α − 0.6·β) / total_power [−1, +1]
Arousal level (β + γ) / total_power [0, 1]

See docs/bci_math.md for full derivations and ICA pipeline notes.


ECDSA Provenance

Signing (FIPS 186-5 / SEC 1 v2.0 - secp256k1)

Step Operation
1 canonical_json = serde_json::to_string(InferenceResult)
2 e = SHA-256(canonical_json)
3 Sample ephemeral k from CSPRNG
4 R = k·G; r = R.x mod n
5 s = k⁻¹(e + r·d) mod n
6 Signature = r‖s, 64 bytes → hex-encoded in SignedOutput.signature_hex

Offline verification

// No private key required - public key embedded in SignedOutput.
let valid = EcdsaSigner::verify_signed(&signed_output)?;

// Or use the standalone verifier:
let verifier = EcdsaVerifier::from_hex(&signed_output.public_key_hex)?;
verifier.verify(&sha256_hash, &signed_output.signature_hex)?;

Build & Run

Prerequisites

Tool Version Notes
Rust stable ≥ 1.85.0 (MSRV) rustup update stable
curl any for non-ai-agent live inference
ANTHROPIC_API_KEY - optional; demo mode if absent

Quick start

git clone https://github.com/murtazaai/polar-bear-biochip
cd polar-bear-biochip
cp .env.example .env          # fill in ANTHROPIC_API_KEY for live inference

# Demo mode - no API key required
cargo run -- run --demo --cycles 5

# Live inference
cargo run --release -- run --cycles 10

# With rig-core backend
cargo run --release --features ai-agent -- run --cycles 5

# Verify a signed output file offline
cargo run -- verify signed_outputs/cycle_001.json

All CLI options

USAGE:
  polar-bear-biochip run    [OPTIONS]
  polar-bear-biochip verify <FILE>

run options:
  -c, --cycles <N>        Inference cycles (0 = infinite)  [default: 5]
  -d, --demo              Force demo mode (no API call)
  -o, --output-dir <DIR>  Signed JSON output directory     [default: signed_outputs]
  -m, --model <MODEL>     Anthropic model                  [default: claude-sonnet-4-6]

Examples

cargo run --example sensors_demo           # sensor fusion table (no key)
cargo run --example provenance_demo        # ECDSA sign/verify/tamper (no key)
cargo run --example agent_demo             # agent demo (demo fallback)
cargo run --example agent_demo --features ai-agent   # live rig-core inference

Sample output

  ╔═══════════════════════════════════════════════════╗
  ║   Polar Bear  ·  Bio-Chip Intelligence Framework  ║
  ║   Sensor Fusion  ·  rig-core  ·  ECDSA Provenance ║
  ╚═══════════════════════════════════════════════════╝

INFO  Generating ECDSA keypair (secp256k1)...
INFO  Public Key : 04d23257ac0b...261c583a6ad7

  ──────────────────────────────────────────────────────
  CYCLE 001/005  |  2026-05-19T10:30:00.114Z
  ──────────────────────────────────────────────────────
  [SENSORS] BCI     α=11.5  β=19.2  θ=6.2  δ=2.3  γ=43.9 Hz
  [SENSORS] Accel   x=-0.05  y=+0.23  z=9.69 m/s²  |  Stationary
  [SENSORS] Fused   cogLoad=0.51  valence=-0.00  arousal=0.75  attn=0.62
  [AGENT]   Querying rig-core LLM agent...
  [RESULT]  Alert         : ✅ Normal
  [RESULT]  Cognitive State: Balanced beta-alpha profile - focused, productive engagement
  [RESULT]  Recommendations:
              • Readings within optimal range - maintain current activity
              • Beta dominance confirms active problem-solving mode
              • Schedule a 5-min micro-break within 45 minutes
  [PROV]    Hash      : 1d588fa83a0d8d1a9ffb...
  [PROV]    Signature : a304cd96500e3c4c4a5a...
  [PROV]    ✓  ECDSA signature verified inline (secp256k1 / SHA-256)
  [PROV]    ✓  Signed output → signed_outputs/cycle_001.json

Test inventory

Unit tests (inline in source modules)

Module Tests Coverage
sensors/bci.rs 3 Band ranges, derived indices, Default
sensors/accelerometer.rs 4 Gravity axis, magnitude, activity states, Default
sensors/fusion.rs 3 Feature bounds, sequence_id, Default
agent/biochip_agent.rs 3 Demo scenarios, Default, prompt builder
provenance/ecdsa_signer.rs 7 Sign/verify, tamper, from_hex, verifier, geometry

Integration tests (tests/)

File Tests Coverage
tests/sensor_tests.rs 12 BCI bands, accel geometry, fusion boundaries, timestamps
tests/provenance_tests.rs 14 ECDSA geometry, sign/verify, tamper detection, key ops

Total: 46 tests

cargo test          # run all 46 tests
cargo test sensors  # sensor layer only
cargo test provenan # provenance layer only

Signed output format

{
  "inference_result": {
    "timestamp": "2026-05-19T10:30:00.117Z",
    "sequence_id": 1,
    "fused_reading": {
      "bci": { "alpha_hz": 11.5, "beta_hz": 19.2, "attention_index": 0.62, ... },
      "accelerometer": { "x": -0.05, "z": 9.69, "activity_state": "Stationary", ... },
      "cognitive_load": 0.51,
      "emotional_valence": -0.0,
      "arousal_level": 0.75
    },
    "cognitive_state": "Balanced beta-alpha profile - focused engagement",
    "alert_level": "Normal",
    "recommendations": [ "...", "...", "..." ]
  },
  "payload_hash_hex": "1d588fa83a0d8d1a9ffb...",
  "signature_hex":    "a304cd96500e3c4c4a5a...",
  "public_key_hex":   "04d23257ac0b...",
  "signed_at":        "2026-05-19T10:30:00.117Z"
}

Tamper any field → SHA-256 hash changes → ECDSA signature invalidated.


rig-core integration

// Without --features ai-agent (curl subprocess - same JSON payload as rig-core):
let agent = BioChipAgent::new("claude-sonnet-4-6", demo);
let result = agent.infer(fused).await?;

// With --features ai-agent (rig-core 0.37 backend - drop-in swap):
// src/agent/biochip_agent.rs rig_inference():
let client = anthropic::Client::from_env()?;
let agent  = client.agent(model).preamble(PREAMBLE).max_tokens(512).build();
agent.prompt(build_prompt(reading)).await

Both paths produce identical InferenceResult structures.


License

Proprietary - © 2026 Murtaza Ali Imtiaz / Polar Bear Systems
See LICENSE-PBS for permitted use.


Author

Murtaza Ali Imtiaz

About

Bio-chip intelligence - sensor fusion, LLM orchestration and provenance

Resources

License

Contributing

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages