PoW+PoS hybrid consensus, post-quantum crypto (ML-DSA-87 + ML-KEM-1024), and a real CLI wallet

[devnull37]

Summary

Turns the chain into a working hybrid Proof-of-Work + Proof-of-Stake network with real post-quantum cryptography, hardens the consensus pallet, wires real weights, and adds a genuine CLI wallet + PQ tooling.

Block authoring is real PoW via sc-consensus-pow (double-Blake2-256 over pre_hash || nonce, U256 difficulty, on-chain retargeting); Aura/GRANDPA were removed from the node and runtime. pallet-ghost-consensus is the PoS economic layer (staking, stake-weighted validator selection, 40/60 miner/staker reward split, evidence-gated slash-and-burn, validation-timeout recovery).

What's included

• PoW + PoS consensus — node builds, boots, and mines blocks Refactor and Fix Ghost Chain Consensus #1→#20.
• Post-quantum signatures — on-chain ML-DSA-87 (Dilithium-5, FIPS 204) verification in the no_std Wasm runtime; validators register keys, and validate_block enforces an ML-DSA signature from the selected validator when a key is registered.
• Post-quantum encryption — node-side ML-KEM-1024 (FIPS 203) key encapsulation + ChaCha20-Poly1305 AEAD.
• Hardening — replay guard on verify_pq_signature; immediate finality when the staker set is empty (was timeout-only); the mine demo now uses the node's real PoW work function (single source of truth).
• Weights — real operation-grounded SubstrateWeight (DbWeight reads/writes + a conservative ~300µs ML-DSA-verify charge) wired into the runtime, replacing the () placeholder.
• CLI wallet + tooling (ghost <cmd>, over JSON-RPC):
  • balance, validators — live state queries (validators also shows ML-DSA key status).
  • stake, unstake, transfer — build, sr25519-sign, and submit real extrinsics.
  • pq-keygen, register-key — ML-DSA-87 keygen + on-chain registration.
  • pq-kem-keygen, pq-encrypt, pq-decrypt — ML-KEM-1024 file encryption (offline).

Verification

• 37 pallet tests + 20 node tests pass.
• Live --dev run: mined Refactor and Fix Ghost Chain Consensus #1→#20; stake (2→5 UNIT), unstake (5→4 UNIT), and transfer (Bob +5 UNIT) executed with correct nonce sequencing; register-key then showed ML-DSA-87 key registered (2592 bytes) in validators; ML-KEM pq-encrypt→pq-decrypt round-tripped exactly and a wrong-key decrypt was correctly rejected.

Honest scope / caveats

• Node-to-node transport is classical libp2p Noise/X25519 — stable2407 has no PQ-Noise variant. PQ lives at the signature + application-encryption layers, not the wire transport.
• No external security audit has been performed; finality is probabilistic longest-chain PoW, not BFT.
• No supply cap (inflationary subsidy — a deliberate monetary-policy decision to make before mainnet).
• Weights are conservative and operation-grounded; empirical frame-benchmarking numbers are still recommended before mainnet.
• Windows build note: [profile.dev] debug = false is set to avoid the MSVC LNK1318 PDB-size limit on the node binary.

Build & test

cargo build --bin ghost-node
cargo test -p pallet-ghost-consensus   # 37 tests
cargo test -p ghost-node               # 20 tests
cargo run --bin ghost-node -- --dev    # mines real PoW blocks

🤖 Generated with Claude Code

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.