|
PQ-WB-KEM: White-Box ML-KEM-768 1.0
PQ-WB-KEM: Toward a White-Box Construction of ML-KEM-768 with Arithmetic Masking for M2M Communications - Mathematics (MDPI)
|
Entry-determinacy dichotomy for the white-box ML-KEM-768 table families: measures the entry-determinacy degree d and the minimal table-access budget M* at which an access-pattern-oblivious, table-access-bounded adversary attains Z=1. More...
Go to the source code of this file.
Namespaces | |
| namespace | entry_determinacy |
Functions | |
| entry_determinacy.bijection_determinacy (perm) | |
| Entry-determinacy of a Z_q bijection by completion: with one entry omitted the missing image is the unique unused value; with two omitted the completions can be swapped. | |
| entry_determinacy.build_encoding_sublayer (rng) | |
| Output-encoding sublayer: per factor two secret Z_q bijections plus their inverses (only the two forward maps are independent). | |
| entry_determinacy.build_prf_table (key) | |
| SPACE-style control: q-entry table whose entries are AES-128(key, index) truncated into Z_q. | |
| entry_determinacy.linear_attack (table, M, inv) | |
| Read one entry at each candidate index, invert the linear model s = T[x0]*x0^-1, regenerate the whole table and score it. | |
| entry_determinacy.main () | |
| Build both constructions plus the encoding sublayer and the AES-128 PRF control, measure entry-determinacy degrees, M* budgets and the key-determinism separator, and write JSON. | |
| entry_determinacy.modular_inverses () | |
| Multiplicative inverses of 1..q-1 mod q (Fermat; q is prime). | |
Variables | |
| int | entry_determinacy.AES_BLOCK_BYTES = 16 |
| int | entry_determinacy.AES_KEY_BITS = 128 |
| int | entry_determinacy.EXPECTED_D_BAKED = 1 |
| int | entry_determinacy.EXPECTED_D_BIJECTION = Q - 1 |
| int | entry_determinacy.EXPECTED_D_SHARED = 0 |
| int | entry_determinacy.EXPECTED_M_STAR_BAKED_BYTES = 1536 |
| int | entry_determinacy.EXPECTED_M_STAR_SHARED_BYTES = 0 |
| int | entry_determinacy.EXPECTED_PRF_LINEAR_REGENERATIONS = 0 |
| str | entry_determinacy.PRIOR_FOOTPRINT_RANGE_TEXT = "7.59 MB to 17.02 GB" |
| tuple | entry_determinacy.REJECTION_BOUND = (1 << 16) // Q * Q |
| int | entry_determinacy.REJECTION_RETRY_LIMIT = 64 |
Entry-determinacy dichotomy for the white-box ML-KEM-768 table families: measures the entry-determinacy degree d and the minimal table-access budget M* at which an access-pattern-oblivious, table-access-bounded adversary attains Z=1.
The entry-determinacy degree d of a table family is the minimum number of entries that determine the whole table. Two regimes are separated here by actual reconstruction attempts, not by assertion: NEGATIVE (algebraically determined): the budget at which the adversary attains Z=1 is M* = (independent secret scalars) x (bytes per entry), with no S_tbl term. Construction B bakes h(x) = shat*x mod q, which is Z_q-linear: one entry at a nonzero index yields shat by a modular inverse, so d=1 and the whole family regenerates from 2 entries per degree-2 factor (h2 is derived from shat1 and the public gamma_i). Construction A stores the public map M[x][y] = x*y mod q, which carries no secret: d=0 and M*=0. POSITIVE (entry-unpredictable): a random Z_q bijection admits no completion from a proper subset beyond its last entry (d = q-1), and an AES-128 PRF table admits none at all within the tested attack classes. The separator is NOT key-determinism: the SPACE-style PRF table is measured here to be exactly as key-determined as B (rebuilt byte-identically from the same key) yet resists the same attack. The separator is efficient local invertibility – B's key->table map is linear and inverts from one entry; AES-128's is a PRF.
results/entry_determinacy.json - Entry-determinacy degrees, M* budgets, PRF controlDefinition in file entry_determinacy.py.