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)
Loading...
Searching...
No Matches
whitebox_reference.json
Go to the documentation of this file.
1{
2 "seed": 42,
3 "q": 3329,
4 "n": 256,
5 "k": 3,
6 "num_components": 128,
7 "secret_domain": {
8 "coefficient_domain": "CBD, s in {-2..2}",
9 "ntt_domain_operand": "shat = NTT(s), full-range over Z_q",
10 "shat_range": [
11 7,
12 3325
13 ],
14 "cbd_range_occupancy": {
15 "basis": "measured",
16 "definition": "fraction of NTT-domain secret coordinates falling in the five-value coefficient-domain range {-2..2} reduced mod q",
17 "reference_sample": {
18 "seed": 42,
19 "k": 3,
20 "coordinates": 768,
21 "in_cbd_range": 0,
22 "percent": 0.0
23 },
24 "multi_seed": {
25 "seeds": 200,
26 "coordinates": 153600,
27 "in_cbd_range": 226,
28 "percent": 0.1471
29 },
30 "uniform_prediction_percent": 0.1502,
31 "note": "The rate is NOT zero, though this reference sample happens to contain none -- which is exactly why one sample cannot settle it. The measured rate tracks the 5/q rate expected of a coordinate ranging over all of Z_q, which is the point: the NTT image inherits no coefficient-domain smallness. A rate of exactly 0 on a finite sample is a property of that sample, not of the distribution."
32 },
33 "note": "The base multiply operand is the full-range NTT-domain secret; coefficient-domain smallness does not carry into the NTT domain."
34 },
35 "correctness_trials": 5000,
36 "correctness_passed": true,
37 "wholevector_crosscheck_passed": true,
38 "constructions": {
39 "A_shared_full_table": {
40 "description": "One shared q x q multiply table; secret enters as a full-range index; secret-independent, shared across all components and all k polynomials.",
41 "total_bytes": 22164482,
42 "total_MB": 22.16,
43 "overhead_factor": 2.07,
44 "overhead_factor_machine_dependent": true,
45 "basis": "measured (table built and serialized; overhead on simulation host)"
46 },
47 "B_baked_per_component": {
48 "description": "Per-component q-entry tables with the fixed secret baked in (h0,h1,h2), indexed by one full-range ciphertext coordinate; scales with module rank k.",
49 "per_component_bytes": 19974,
50 "per_secret_poly_bytes": 2556672,
51 "per_secret_poly_MB": 2.56,
52 "total_bytes": 7670016,
53 "total_MB": 7.67,
54 "overhead_factor": 1.5,
55 "overhead_factor_machine_dependent": true,
56 "basis": "measured (tables built and serialized; overhead on simulation host)"
57 }
58 },
59 "determinism": {
60 "footprints_deterministic": true,
61 "machine_dependent_fields": [
62 "constructions.A_shared_full_table.overhead_factor",
63 "constructions.B_baked_per_component.overhead_factor"
64 ],
65 "note": "All size/footprint fields are seeded (seed 42) and byte-identical run-to-run; the overhead_factor fields are wall-clock ratios measured on the simulation host and vary run-to-run. run_all.py flags this experiment machine_dependent=False for its deterministic footprints; the byte-identical claim excludes the two overhead_factor timings."
66 },
67 "note": "Reference NumPy/Python measurement on the simulation host. Both constructions operate on the real full-range NTT-domain secret and are verified against the reference base multiply. These are the measured base tables; the full deployed footprint is likewise measured (encoding sublayer in whitebox_encoded.py, coefficient shuffle in obfuscation_storage.py), and only the deployment latency remains an analytical projection."
68}