6 "ntt_domain_operand": "shat = NTT(s), full-range over Z_q",
12 "correctness_trials": 5000,
13 "correctness_passed": true,
15 "scheme": "per-component secret nonlinear bijection on Z_q for each of the two output coordinates (128 x 2 encodings + inverses)",
17 "storage_bits_per_encoding": 39948,
18 "key_entropy_bits_per_encoding_log2_qfact": 34156.7,
19 "basis": "measured (encodings built and serialized)"
22 "encoding_tables_bytes": 3408896,
23 "encoding_tables_MB": 3.41,
24 "note": "encoding tables store the 256 secret bijections and their inverses as uint16 arrays; this is a MEASURED lower bound on the encoding-layer footprint. It acts on the full-range output coordinates and is independent of the base-table construction (A shared / B baked; see whitebox_reference.py)."
27 "components_tested": 128,
28 "queries_per_component": 3,
29 "recovery_plain_unencoded": 128,
30 "recovery_encoded_encoding_unaware": 0,
31 "recovery_encoded_encoding_aware": 128,
32 "interpretation": "On the full-range NTT-domain secret, the plain linear-solve recovers (shat0,shat1) from unencoded intermediates and from encoded intermediates ONLY when the secret encoding is known; against the encoding-unaware (mixing-aware) attacker it fails. This is an EMPIRICAL NEGATIVE result for the tested attacker, not a proof: a structural attacker that recovers the encodings (BGE-style, open for the lattice setting) is not modelled here."