Security
Who could leak a paper, and what stops them. Each row says whether the protection runs on testnet today or is next on the roadmap.
Threat model
| If… | What happens | Status |
|---|---|---|
| Someone at the press, in transport or in a strongroom copies the paper | They only ever hold ciphertext. The readable paper does not exist until the bell. | Live |
| A centre tries to open early | Its key opens only its own envelope, and the time half is missing until 2 keypers post. recordOpening reverts with NotUnlocked. | Live |
| One keyper leaks its seed early | One seed opens nothing. Whoever holds the early seed can commit-reveal it and take half that keyper's bond. This happened on testnet: a bribed keyper was slashed 14 minutes before its exam opened (see the release ledger). | Live |
| Keypers collude | They need the threshold of seeds before the bell, and each accomplice can betray the rest for half a bond. Cost grows with threshold and bond (calculator below). | Live |
| A keyper is offline at the bell | Any threshold of keypers opens it. If the paper itself is compromised, the board switches every centre to a backup paper sealed to a later round. | Live |
| A centre leaks after opening | Every centre's paper is a different permutation. A few questions from a photo name the centre; the Leak Hunter files a signed report. | Live |
| Someone swaps a ciphertext file | Each centre's ciphertext hash is on chain before the exam. The console refuses a file that does not match. | Live |
| The board seals to an attacker's key | Centre keys are read only from QoreChain's KEM registry and bound to the account by a hybrid-signed recipient card. | Live |
| A quantum attacker records everything today | Every lock and signature is post-quantum: ML-KEM-1024, QoreSeal hybrid, ML-DSA-87. | Live |
| The RPC operator sees a seed in an eth_call | Keyper daemons never let a seed leave the machine before chain time passes the round, not even for gas estimation. | Live |
| A keyper's ECDSA key is stolen | Its EVM window is scoped to the Hourglass registry and needs an ML-DSA-87 signature to open. We hand our drill key to the attacker bot on purpose. | Live |
| The exam board itself leaks the paper | The board writes the paper, so ZeroHour cannot hide it from the board. The fairness reveal proves no centre got a different paper. Splitting authorship across setters is next. | Next |
Cost of corruption
Opening a paper early means getting a threshold of keypers to break their one rule, while each of them can be paid to betray the others.
Keypers in the exam: 3
Seeds needed to open (threshold): 2
Bond per keyper: 1 QOR
2 QOR
at risk to open one paper early, across 2 people who each gain half a bond by turning the others in.
Today's testnet exams run 2 of 3 keypers at 1 QOR each. The contract allows any threshold and bond: a national exam would set both far higher and draw keypers from different institutions.
What is real today
| Piece | Status |
|---|---|
| HourglassRegistry and ZeroHourExam on QoreChain testnet | Live on testnet |
| Three bonded keypers with ML-KEM-1024 round keys and daemons | Live on testnet |
| Two centres with on-chain KEM keys and signed recipient cards | Live on testnet |
| In-browser opening with ciphertext, seed and anchor checks | Live on testnet |
| Leak tracing against 1,000 variants (2 real centres, 998 simulated) and signed on-chain reports | Live on testnet |
| Public fairness reveal, recomputed in the browser | Live on testnet |
| Leak Hunter reading a real phone photo with Claude vision | Next |
| Outside keypers from other institutions | Next |
| Multi-setter sealing so no single person holds the paper | Next |
| Least-privilege hot key for keypers | Next |