SECURITY_HARDENING

Security & privacy hardening (MeshChain)

Honest framing

No networked money system is “absolutely bulletproof.”
Anyone who claims perfect security or perfect privacy is selling fiction.

MeshChain’s bar is maximum practical rigor under real constraints (LoRa airtime, RF physics, Solana/BTC classical L1s):

• Defense in depth — many independent locks must all fail
• Fail secure — if unsure, reject
• Least privilege — internet path cannot unlock without mesh proof
• Cryptographic authenticity — spends require keys; large/cold paths use ML-DSA-65
• Privacy by design — minimize linkability; never put seeds or raw bank identifiers on air
• Explicit residual risk — operators and users know what remains

If a property is not proven, we do not claim it.

---

Target properties

Security (integrity & custody)

Property	Mechanism
Only owner spends	ed25519 + optional ML-DSA-65 on body
No double-spend	per-account nonces + finality
No silent mint	authorized minter set only
Vault not freeable via internet alone	hybrid lock: mesh_short_id + burn_txid + K attestors
Replay unlock	unique burn_txid PDA; deposit amount_unlocked cap
Quantum-resistant cold auth	ML-DSA-65 (FIPS 204) for large/vault burns
Censorship reduced	multi-validator PoA ceil(2N/3)
Fail secure hybrid	hybrid forced on production configs; min attestors ≥ 2

Privacy (confidentiality & unlinkability)

Property	Mechanism	Limit
No KYC in protocol	keys only	Social graph still exists
Pseudonymous addresses	short ids from key hashes	RF can still localize radios
No plaintext redeem address on mesh	redeem_hint = hash only	Bridge learns dest when unlocking
One-time receive tags	stealth short-id derivation	Optional; user must use it
No seeds on mesh/MQTT/logs	policy + code review	Operator mistakes
Channel isolation	private Meshtastic channel for funds	PSK compromise → metadata

RF is not Tor. Direction-finding, timing, and traffic volume are residual.

---

Hybrid dual-control (bullet-resistant vault)

Unlock requires ALL:
  [1] deposit.mesh_short_id == provided mesh_short_id
  [2] burn_txid unique + non-zero (mesh Burn happened)
  [3] amount ≤ claim remaining
  [4] ≥ K registered mesh attestors co-sign the Solana tx
  [5] program not paused; vault solvent

Internet-only (hacked website, stolen relayer hot key, compromised RPC) → insufficient.

---

Cryptographic suite

Use	Algorithm	Notes
Everyday small mesh tx	ed25519	Small packets; not long-term PQ
Cold / large / vault burn	ML-DSA-65	Multi-FRAG on LoRa
Hashing	SHA-256 with domain separation prefixes
Mesh framing	MC magic + typed ports	Reject unknown
Vault L1	Solana program constraints	L1 still classical

Domain prefixes used in code:

• meshchain-v2-mldsa65 — PQ context
• meshchain-pq-id — PQ short id
• meshchain-redeem-v1 — redeem destination hash
• meshchain-stealth-v1 — one-time receive id

---

Operational hard rules

1. Cold keys never on phones with SIM/5G or always-on cloud backup.
2. Attestor keys geo-separated; threshold K ≥ 2; never all on one VPS.
3. Meshtastic funds channel PSK ≠ custody; rotate if leaked.
4. Power radio only for intentional sessions.
5. Prefer fresh Solana/BTC destination on every unlock.
6. No balance/identity publishing to public MQTT by default.
7. Audit set_attestors / authority like root.

---

Residual risks (must accept or mitigate off-protocol)

Risk	Severity	Mitigation
≥K attestors collude	Critical	Diversify operators; social monitoring
≥2/3 mesh validators censor/halt	High	Diverse validators; user exits while healthy
RF traffic analysis	Medium	Rare sessions; cover traffic sparingly
Solana/BTC L1 quantum break (future)	High long-term	Migrate vaults when L1s upgrade; minimize time locked
User loses cold seed	Critical	Metal backup; optional social recovery (out of band)
Implementation bugs	Critical	Tests, review, formal audit before mainnet value
Supply-chain / malware on build host	Critical	Reproducible builds; air-gapped sign

---

Before real value

• Independent security review of vault + ledger
• Hybrid attestors ≥ 3 with documented ops
• PQ required for all vault-related burns
• Private mesh channel + RF hygiene guide
• Incident response: pause bridge, freeze mint
• No mainnet without bug bounty / staged limits

Bottom line: We engineer for hostile internet and curious RF adversaries. We do not claim magical invulnerability.