The package was missing every script behind the KM (lrn.) curves and
both learned table rows: exp_learned's driver, the merge that folds the
learned rows into sec_compare.csv and refresh_summary.csv, and the
report that reads the learned numbers back. It was also missing
sec_keylen_perm.csv, so Fig. 3 could not be regenerated at all, and the
two diagnostics that answer why a fixed key beats a learned one here and
where a learned mask would win instead.
The learned artifacts themselves are regenerated. They were trained on
the cross-entropy alone, which drifts to disjoint sparse supports: 99
percent of each key's energy on about six of the 64 entries, so a digit
is decided over a sixth of its period and the key set is a choice of
support rather than a dense direction in R^L. They are now the
regularized keys of Section V-C, and check_consistency asserts which of
the two families the figures draw.
Verified from inside this repository: replot_security.py rebuilds all
seven result figures, make_tables.py reproduces both result tables, and
check_consistency.py passes every check that does not need the
manuscript.
The README now lists what ships. Its run list, layout and figure map had
none of the learned pipeline, named two tables the manuscript renders as
prose, and gave Fig. 3 no data file for its permutation curve.
check_family_enum.py now runs both attacks against both families. The
outsider ranks the L-1 Walsh rows; the insider, holding m_v, ranks the
L-1 products m_v .* m_r, which works because Walsh rows are closed
under the elementwise product and the per-block sign cancels in
m_u .* m_v. Both need a list to rank, and only the structured family
supplies one: the structured family falls at 0.905 from one frame at
10 dB and 0.990 from four, the refresh takes the outsider to 0.000 and
leaves the insider at 0.980, and the learned family gives 0.000
throughout.
exp_full.stage_N sweeps the learned family over the same SNR grid at
the same frame count as stage_A, so Fig. 2 can carry both families and
a reader can see what the key space costs at every SNR rather than at
one point.
check_consistency.py gains four assertions for the key-space
measurements and two for the learned sweep, 82 in all.
README: the assertion count was two rounds stale, and the figure map
omitted family_enum, cov_attack and maskdegen, whose CSVs back quoted
manuscript numbers.
check_family_enum.py measures the attack the manuscript now states in
Section III-A: the winning correlation is an index-free verifier, so
ranking the 63 non-constant Walsh rows by mean winning correlation
recovers the user set from one frame in 0.905 of 200 trials at 10 dB
and from four frames in 0.990, using nothing outside the stated threat
model. Under the invariance refresh it recovers it in none, because the
entry permutation relabels the codebook the adversary must align
against.
V8 and V9 read the trained codebook through main_model(), which
retrains on every call, and a codebook trained on CUDA is not the one
trained on CPU. The shipped verify_math.csv therefore read PASS here
and FAIL for anyone running this package without a GPU. model_main.pt
is 7 KB and fixes the codebook, which is what both checks are about;
delete it to retrain. V1-V11 now pass on both.
New checks: V10, the format-matched OMA reference Section VI-B quotes,
and V11, the closed-form against Monte Carlo comparison the manuscript
claimed and never stored. V3a's bias-linearity result was computed and
printed but never written to the CSV, so the one linearity claim the
paper quotes was the one this package could not show.
check_consistency.py gains 21 assertions, covering five data files that
no assertion read (users, csi, semantic, cov_attack, sec_jam) and the
trend claims it structurally could not see, since it compared values
and not shapes.
README: the figure map named stages that do not write the artifacts
they list, so following it did not reproduce Figs. 4 and 6; the
reproduction block was five scripts short; and the refresh numbers were
from a superseded run (nearly three, 15.0 to 64.8 bits) against the
manuscript's 2.3 and 23.8 to 364.6.
The jammed OMA model now matches the transmit chain the paper
describes, the key-length sweep averages the eavesdropper over eight
substitute-key draws, and verify_math gains the coded-OMA outage
reference, symbolic checks of the three algebraic identities, and the
cross-period remainder of Proposition 2. The README records the energy
and channel conventions.
Resource-match the OMA reference in the key-length sweep (oma_ser_keylen),
which gives it the L/16 combining gain the longer frame allows. The
proposal now passes a resource-matched OMA by 1.27x at L=64 rather than
the 4.3x reported against a fixed-d reference.
Densify the JSR, sensitivity, and brute-force grids so the curves are
smooth, give the index cipher its channel floor instead of error-free
reception, and add the permutation-key known-plaintext attack
(exp_permkpa) so Fig. 7 carries a conventional linear scheme.
Add check_cov_attack.py and check_cov_ceiling.py: a referee raised a
ciphertext-only second-order attack; the exact-population test shows the
received covariance leaks only a sparse rank-deficient subset of the key
Gram and leaves the eavesdropper at the random-guess level.
Dump verify_math.csv, move the superseded V=256 pilot CSVs to data/pilot.
Give the orthogonal-access jammer its own Rayleigh channel in
oma_ser_jammed, matching the convention every simulated scheme already
used. Without it the closed-form curve faced a jammer at full power in
every frame while the Monte Carlo curves faced a fading one, which
inverted the ordering of the comparison.
Measure the outsider error rate for the fixed-key and naive-refresh
cases as well, and emit the two refresh tables from make_tables.py, so
no cell of the paper is hand-typed.
Derives each block key from the transformations that leave the decision
statistic invariant: a global sign per frame entry, a relabeling of the
frame entries applied to keys and codebook together, and a permutation
of which user holds which row. The legitimate error rate is unchanged
at 0.257 across 24 blocks while a refresh that installs fresh
orthogonal keys reaches 0.710, and a key recovered by known plaintext
returns to the random-guess level one block later.
Reproducibility package for the TIFS submission: transmit and receive
core, security stages (eavesdropper, jamming, key families, attack
difficulty, known-plaintext), real BERT token streams, closed-form
verification, and the scripts that regenerate every figure and table
from the released CSVs.
KiHoLee