PairFinite
The entanglement-based finite-key length of Tomamichel & Leverrier 2017
16/16 passed in 1382ms
| Test | What it does | Result |
|---|---|---|
test_asymmetry | The refusal states the two-sided gap as a requirement, naming Alice's side as the half that does not transfer. | ✅ pass |
test_asymptote | The rate per sifted round rises with the block towards log2(1/cbar) - 2h(delta), reaching 98% of it at 1e12. | ✅ pass |
test_budget | Every length pair_finite returns costs no more than the epsilon it was asked for, read back through pair_secpar. | ✅ pass |
test_budget_domain | A security target at or above one half is refused, at pair_finite and at pair_length's eps_pa. | ✅ pass |
test_cbar | Alice's complementarity enters the length as log2(1/cbar) per key round, and cbar = 1 is refused. | ✅ pass |
test_coincidence | A coincidence-counted pair link is refused by name, the message naming three missing pieces and four papers. | ✅ pass |
test_fuzz | Across 150 random blocks, error rates, complementarities and targets, no length pair_finite returns overspends its epsilon. | ✅ pass |
test_inverse | pair_length and pair_secpar invert each other to within the one bit the floor gives away. | ✅ pass |
test_marker | At each block length Figure 7 marks, the length per sifted round has just reached half the asymptotic 1 - 2h(delta). | ✅ pass |
test_optimised | pair_finite at those points is never shorter than the published length and at most a part in a thousand longer. | ✅ pass |
test_published | At the four block lengths Tomamichel and Leverrier's Figure 7 marks, pair_length and pair_secpar return the length and the epsilon their own code printed. | ✅ pass |
test_short | A block too short for its security parameter raises rather than returning zero, and an unknown detection model is refused. | ✅ pass |
test_slots | Out-of-domain arguments to the finite-key entry points raise rather than returning a wrong number. | ✅ pass |
test_squasher | The refusal does not blame the squashing model, naming the three Gittsovich theorems, the vacuum flag and Eq. (134) instead. | ✅ pass |
test_terms | The parameter-estimation and correctness terms are Eq. (58)'s exponential and Theorem 2's 2^-t, and the total is their sum. | ✅ pass |
test_vanishes | The length falls to nothing exactly where log2(1/cbar) meets 2h(delta). | ✅ pass |