# 6G Native ISAC & the Sovereign Tensor Network

**Organization:** Milkyway Center | **Contact:** yuanjimaster@gmail.com | **Site:** https://milkyway.center

A **sovereign physical layer** for the 6G era: one coherent-optical waveform that carries communication *and* nine-dimensional physical sensing at once — verified on real silicon — running on its own link and protocol, free of 3GPP standard-essential patents (SEPs) and spectrum licensing.

**Two harvests per device. One field per network.** Every device decodes communication and a 3×3 physical-property tensor from the *same* received capture, and writes those measurements into one shared **Tensor Potential Field (TPF)** — so every participant shares the same physical reality.

## Why "native", and why sovereign

Conventional ISAC splits one waveform by time-slots and pilot insertion — communication and radar fight for the same resource. V131 is different at the root: transmitter and receiver read the **same waveform ledger** (the S01 physical anchor — "what we emit is what we accept"), so sensing and communication are born from one physical identity, not synchronized after the fact. Because the medium is **coherent optical**, not licensed radio, the RF standard-essential patents that gate the cellular world do not read on it, and no spectrum auction gates deployment. The network is not a tenant in anyone's building.

## Verified — Native ISAC on silicon

| Claim | Result | Status |
|-------|--------|--------|
| Single capture, dual decode | one SHA-256-pinned capture feeds **both** decoders | Hardware (FPGA) |
| Communication path | **1,720 blocks / 10,567,680 bits, zero bit errors** | Hardware |
| Sensing path | nine columns full-rank on the same capture, fidelity ≥ design | Hardware |

The two decoders consume the **same original capture file** (pinned by SHA-256) — not two separate acquisitions. That single-capture, dual-decode property, verified through an FPGA link at **zero bit error over 10.57 million bits**, is the electrical-level proof that "one waveform, two harvests" is real, not a marketing composite.

## Verified — information-theoretic confidentiality

Security here is a pre-registered, falsifiable result. **Threat model (Kerckhoffs):** the adversary knows the entire system, captures the full ciphertext, and may hold known plaintext — but does **not** hold the key (the S01-derived one-time material). Under that model, the keyed construction (one-time pad + encrypt-then-MAC) is verified to deliver:

| Property | Measured | Grade |
|----------|----------|-------|
| Confidentiality | ciphertext independent of plaintext; adversary recovery = blind guess (0.4986) | Shannon info-theoretic |
| Tamper-evidence | tampering & forgery detected 100%; forgery without the key ≈ 2^-256 | verified |
| Key discipline | known plaintext on one message leaks nothing about an independent message | verified |

Two deliberately-broken **canary** constructions — key reuse (two-time pad) and integrity stripped — are correctly caught and **fail** the criterion, proving the judge has teeth.

**What "absolute" honestly means here.** This is Shannon *information-theoretic* perfect secrecy — I(ciphertext; plaintext) = 0, so brute force does not help — but it is absolute **only inside its stated model**, exactly as the one-time pad has been provably unbreakable for eighty years *given its assumptions*. The guarantee carries its price and preconditions:

- **Three preconditions:** the key is truly random, kept secret, and used once; and ciphertext integrity is protected by the MAC. Break any one and the guarantee is void.
- **Capacity price (Shannon's price):** the secure payload cannot exceed the key material — measured in **time samples of the ψ(t) waveform**, the physical one-time pad. Beyond that, one re-keys or drops to computational-grade security.

We therefore do **not** publish "unconditionally absolute security." We publish something stronger because it is true: *perfect secrecy under an explicit threat model, with its preconditions and capacity stated.*

## The Tensor Potential Field — one field per network

Per-device ISAC is the starting point; the network form is the **Tensor Potential Field**. Every device writes its nine-dimensional measurements into one shared field and reads back the risk-weighted cost of every action — coordination stops being a negotiation protocol and emerges in the field. *(Field-scale results are simulation, reproducible on GPU.)*

- **Measure once, reuse network-wide:** multi-device fusion +5.6 dB — one node detects a hazard, the network reroutes before arrival.
- **10k-scale airspace:** 16,384 drones coordinated online in one airspace; one hazard broadcast reroutes the network in a single step.
- **No cooperation required:** non-participants are written in as measured objects — penetration rate falls from a life-or-death threshold to a density parameter.
- **Poisoning does not break it:** injected false data — 100% detected, 0% false accusations.

## Honest boundaries

- ISAC is verified at the **electrical / digital level** (FPGA link); the 1550 nm coherent optical unit is **in build**, criteria pre-registered.
- The confidentiality result is at the **protocol / data layer**, with a surrogate key standing in for the master. Physical-layer interception-evidence (a tap that measurably disturbs the coherent state) is a **research axis, not yet verified** — it awaits the optical bench.
- Tensor-field results (fusion gain, 16,384-drone coordination, poisoning defense) are **simulation**.
- The patent posture is our engineering assessment, not legal advice; a formal freedom-to-operate opinion is separate work.

Every claim above is bound to a pre-registered, falsifiable criterion — locked before the run, allowed to fail, failures filed as-is. Same stance as the rest of V131: run toward the tests, not away from them.