Newer
Older
#!/usr/bin/env python
from functools import cmp_to_key
import os
import sys
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import scipy
import helper_scripts.helper_functions as helper_functions
import helper_scripts.performance_eval_of_oqs as performance_eval_of_oqs
RESULTS_DIR = "saved/results-run-20250207-vm-p16"
FILTER_RESULTS = []
PLOTS_DIR = "plots"
FEATHERS_DIR = "feathers"
cmap = plt.cm.hsv
# cmap = plt.colormaps.get_cmap("nipy_spectral")
def main():
data = load_data()
# generally they both only seconds for graphs when not generating for single algorithms
plot_general_plots() # takes about 4 seconds
plot_lines(data) # takes about 1:50 min
plot_static_data(data) # takes about 4 min
def load_data():
if os.path.exists(f"{FEATHERS_DIR}/data.feather"):
data = pd.read_feather(f"{FEATHERS_DIR}/data.feather")
else:
data = read_data_into_pandas()
return data
# Reading in the data takes about 11 seconds per scenario
def read_data_into_pandas():
data = pd.DataFrame(
columns=[
"scenario",
"protocol",
"sec_level",
"kem_alg",
"srv_pkt_loss",
"srv_delay",
"srv_jitter",
"srv_duplicate",
"srv_corrupt",
"srv_reorder",
"srv_rate",
"cli_pkt_loss",
"cli_delay",
"cli_jitter",
"cli_duplicate",
"cli_corrupt",
"cli_reorder",
"cli_rate",
"error_count",
"error_rate",
"measurements",
"mean",
"std",
"cv",
"median",
"qtl_25",
"qtl_75",
"qtl_95",
"qtl_99",
"iqr",
"skewness",
"kurtosis",
]
)
def get_all_result_files():
result_files = []
for dirpath, _, filenames in os.walk(RESULTS_DIR):
if filenames and not any(
filter_value in dirpath for filter_value in FILTER_RESULTS
):
for filename in filenames:
result_files.append(os.path.join(dirpath, filename))
return result_files
for csv_result_file_name in get_all_result_files():
*_, scenario, protocol, sec_level, kem_alg = csv_result_file_name.split("/")
kem_alg = kem_alg.split(".")[0]
# print(f"csv_result_file_name: {csv_result_file_name}")
result_file_data = pd.read_csv(csv_result_file_name, header=None)
result_file_data = result_file_data.T
df_scenariofile = pd.read_csv(f"testscenarios/scenario_{scenario}.csv")
df_scenariofile = df_scenariofile.drop(
df_scenariofile.columns[0], axis="columns"
)
assert len(result_file_data.columns) == len(df_scenariofile)
for i in range(len(result_file_data.columns)):
measurements_and_error_count = result_file_data.iloc[:, i].tolist()
error_count = measurements_and_error_count[0]
measurements = np.array(measurements_and_error_count[1:])
data.loc[len(data)] = {
"scenario": scenario,
"protocol": protocol,
"sec_level": sec_level,
"kem_alg": kem_alg,
"srv_pkt_loss": df_scenariofile.iloc[i]["srv_pkt_loss"],
"srv_delay": df_scenariofile.iloc[i]["srv_delay"],
"srv_jitter": df_scenariofile.iloc[i]["srv_jitter"],
"srv_duplicate": df_scenariofile.iloc[i]["srv_duplicate"],
"srv_corrupt": df_scenariofile.iloc[i]["srv_corrupt"],
"srv_reorder": df_scenariofile.iloc[i]["srv_reorder"],
"srv_rate": df_scenariofile.iloc[i]["srv_rate"],
"cli_pkt_loss": df_scenariofile.iloc[i]["cli_pkt_loss"],
"cli_delay": df_scenariofile.iloc[i]["cli_delay"],
"cli_jitter": df_scenariofile.iloc[i]["cli_jitter"],
"cli_duplicate": df_scenariofile.iloc[i]["cli_duplicate"],
"cli_corrupt": df_scenariofile.iloc[i]["cli_corrupt"],
"cli_reorder": df_scenariofile.iloc[i]["cli_reorder"],
"cli_rate": df_scenariofile.iloc[i]["cli_rate"],
"error_count": error_count,
"error_rate": error_count / len(measurements),
"measurements": measurements,
"mean": np.mean(measurements),
"std": np.std(measurements),
"cv": np.std(measurements) / np.mean(measurements),
"median": np.median(measurements),
"qtl_01": np.quantile(measurements, 0.01),
"qtl_05": np.quantile(measurements, 0.05),
"qtl_25": np.quantile(measurements, 0.25),
"qtl_75": np.quantile(measurements, 0.75),
"qtl_95": np.quantile(measurements, 0.95),
"qtl_99": np.quantile(measurements, 0.99),
"iqr": scipy.stats.iqr(measurements),
"riqr": scipy.stats.iqr(measurements) / np.median(measurements),
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
"skewness": scipy.stats.skew(measurements),
"kurtosis": scipy.stats.kurtosis(measurements),
}
dtypes = {
"scenario": "category",
"protocol": "category",
"sec_level": "category",
"kem_alg": "category",
}
data = data.astype(dtypes)
categories = [
"secp256r1",
"secp384r1",
"secp521r1",
"x25519",
"x448",
"mlkem512",
"p256_mlkem512",
"x25519_mlkem512",
"mlkem768",
"p384_mlkem768",
"x448_mlkem768",
"x25519_mlkem768",
"p256_mlkem768",
"mlkem1024",
"p521_mlkem1024",
"p384_mlkem1024",
"bikel1",
"p256_bikel1",
"x25519_bikel1",
"bikel3",
"p384_bikel3",
"x448_bikel3",
"bikel5",
"p521_bikel5",
"hqc128",
"p256_hqc128",
"x25519_hqc128",
"hqc192",
"p384_hqc192",
"x448_hqc192",
"hqc256",
"p521_hqc256",
"frodo640aes",
"p256_frodo640aes",
"x25519_frodo640aes",
"frodo640shake",
"p256_frodo640shake",
"x25519_frodo640shake",
"frodo976aes",
"p384_frodo976aes",
"x448_frodo976aes",
"frodo976shake",
"p384_frodo976shake",
"x448_frodo976shake",
"frodo1344aes",
"p521_frodo1344aes",
"frodo1344shake",
"p521_frodo1344shake",
]
data["kem_alg"] = pd.Categorical(
data["kem_alg"], categories=categories, ordered=True
)
print(data.head())
print(data.describe())
print(data.info())
print()
print("Scenarios read:", data["scenario"].unique())
os.makedirs(FEATHERS_DIR, mode=0o777, exist_ok=True)
data.to_feather(f"{FEATHERS_DIR}/data.feather")
print("Data written to feather file")
return data
def filter_data(
data,
scenario: str | None = None,
protocol: str | None = None,
sec_level: str | list[str] | None = None,
kem_alg: str | None = None,
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
):
filtered_data = data
# print(filtered_data["kem_alg"] == "x25519") # is a boolean series
if scenario is not None:
filtered_data = filtered_data[filtered_data["scenario"] == scenario]
if protocol is not None:
filtered_data = filtered_data[filtered_data["protocol"] == protocol]
if sec_level is not None:
if type(sec_level) == list:
filtered_data = filtered_data[filtered_data["sec_level"].isin(sec_level)]
else:
filtered_data = filtered_data[filtered_data["sec_level"] == sec_level]
if kem_alg is not None:
filtered_data = filtered_data[filtered_data["kem_alg"] == kem_alg]
def drop_columns_with_only_zero_values(data):
# this complicated way is necessary, because measurements is a list of values
filtered_data_without_measurements = data.drop(columns=["measurements"])
zero_columns_to_drop = (filtered_data_without_measurements != 0).any()
zero_columns_to_drop = [
col
for col in filtered_data_without_measurements.columns
if not zero_columns_to_drop[col]
]
return data.drop(columns=zero_columns_to_drop)
if drop_zero_columns and "measurements" in data.columns:
filtered_data = drop_columns_with_only_zero_values(filtered_data)
# if drop_zero_columns:
# filtered_data = drop_columns_with_only_zero_values(filtered_data)
# print(filtered_data["measurements"].head())
# print(filtered_data)
return filtered_data
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
def get_x_axis_column_name(scenario: str) -> str:
match scenario:
case "duplicate":
return "srv_duplicate"
case "packetloss":
return "srv_pkt_loss"
case "delay":
return "srv_delay"
case "jitter_delay20ms":
return "srv_jitter"
case "corrupt":
return "srv_corrupt"
case "reorder":
return "srv_reorder"
case "rate_both":
return "srv_rate"
case "rate_client":
return "cli_rate"
case "rate_server":
return "srv_rate"
case "static":
assert False, "static scenario has no x-axis"
case _:
print(f"NO MATCH FOUND FOR {scenario}", file=sys.stderr)
sys.exit(1)
def get_x_axis(scenario, data, length):
match scenario:
case "duplicate":
return data["srv_duplicate"]
case "packetloss":
return data["srv_pkt_loss"]
case "delay":
return data["srv_delay"]
case "jitter_delay20ms":
return data["srv_jitter"]
case "corrupt":
return data["srv_corrupt"]
case "reorder":
return data["srv_reorder"]
case "rate_both":
return data["srv_rate"]
case "rate_client":
return data["cli_rate"]
case "rate_server":
return data["srv_rate"]
case "static":
case _:
print(f"NO MATCH FOUND FOR {scenario}", file=sys.stderr)
sys.exit(1)
def map_security_level_hybrid_together(sec_level: str):
match sec_level:
case "secLevel1":
return ["secLevel1", "secLevel1_hybrid"]
case "secLevel3":
return ["secLevel3", "secLevel3_hybrid"]
case "secLevel5":
return ["secLevel5", "secLevel5_hybrid"]
case "miscLevel":
return "miscLevel"
case _:
return None
def get_color_and_mode(kem_alg: str, combined_with_hybrids: bool = False):
# NOTE maybe just use hle colors directly from the start
primary_mode = "-"
secondary_mode = "--" if combined_with_hybrids else "-"
tertiary_mode = ":" if combined_with_hybrids else "--"
secondary_lightness_factor = 0.9 if combined_with_hybrids else 1
tertiary_lightness_factor = 0.8 if combined_with_hybrids else 0.9
a = 0.8 if combined_with_hybrids else 1
no_algos = 8
match kem_alg:
case "secp256r1":
return cmap(0 / no_algos), primary_mode
case "secp384r1":
return cmap(0.3 / no_algos), primary_mode
case "secp521r1":
return cmap(0.6 / no_algos), primary_mode
case "x25519":
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
case "mlkem512":
return cmap(1 / no_algos), primary_mode
case "p256_mlkem512":
return (
transform_cmap_color(
cmap(1 / no_algos),
alpha_factor=a,
lightness_factor=secondary_lightness_factor,
),
secondary_mode,
)
case "x25519_mlkem512":
return (
transform_cmap_color(
cmap(1 / no_algos),
alpha_factor=a,
lightness_factor=tertiary_lightness_factor,
),
tertiary_mode,
)
case "mlkem768":
return cmap(1.3 / no_algos), primary_mode
case "p384_mlkem768":
return (
transform_cmap_color(
cmap(1.3 / no_algos),
alpha_factor=a,
lightness_factor=secondary_lightness_factor,
),
secondary_mode,
)
case "x448_mlkem768":
return (
transform_cmap_color(
cmap(1.3 / no_algos),
alpha_factor=a,
lightness_factor=tertiary_lightness_factor,
),
tertiary_mode,
)
case "mlkem1024":
return cmap(1.6 / no_algos), primary_mode
case "p521_mlkem1024":
return (
transform_cmap_color(
cmap(1.6 / no_algos),
alpha_factor=a,
lightness_factor=secondary_lightness_factor,
),
secondary_mode,
)
case "bikel1":
return cmap(2 / no_algos), primary_mode
case "p256_bikel1":
return (
transform_cmap_color(
cmap(2 / no_algos),
alpha_factor=a,
lightness_factor=secondary_lightness_factor,
),
secondary_mode,
)
case "x25519_bikel1":
return (
transform_cmap_color(
cmap(2 / no_algos),
alpha_factor=a,
lightness_factor=tertiary_lightness_factor,
),
tertiary_mode,
)
case "bikel3":
return cmap(2.3 / no_algos), primary_mode
case "p384_bikel3":
return (
transform_cmap_color(
cmap(2.3 / no_algos),
alpha_factor=a,
lightness_factor=secondary_lightness_factor,
),
secondary_mode,
)
case "x448_bikel3":
return (
transform_cmap_color(
cmap(2.3 / no_algos),
alpha_factor=a,
lightness_factor=tertiary_lightness_factor,
),
tertiary_mode,
)
case "bikel5":
return cmap(2.6 / no_algos), primary_mode
case "p521_bikel5":
return (
transform_cmap_color(
cmap(2.6 / no_algos),
alpha_factor=a,
lightness_factor=secondary_lightness_factor,
),
secondary_mode,
)
case "hqc128":
return cmap(4 / no_algos), primary_mode
case "p256_hqc128":
return (
transform_cmap_color(
cmap(4 / no_algos),
alpha_factor=a,
lightness_factor=secondary_lightness_factor,
),
secondary_mode,
)
case "x25519_hqc128":
return (
transform_cmap_color(
cmap(4 / no_algos),
alpha_factor=a,
lightness_factor=tertiary_lightness_factor,
),
tertiary_mode,
)
case "hqc192":
return cmap(4.3 / no_algos), primary_mode
case "p384_hqc192":
return (
transform_cmap_color(
cmap(4.3 / no_algos),
alpha_factor=a,
lightness_factor=secondary_lightness_factor,
),
secondary_mode,
)
case "x448_hqc192":
return (
transform_cmap_color(
cmap(4.3 / no_algos),
alpha_factor=a,
lightness_factor=tertiary_lightness_factor,
),
tertiary_mode,
)
case "hqc256":
return cmap(4.6 / no_algos), primary_mode
case "p521_hqc256":
return (
transform_cmap_color(
cmap(4.6 / no_algos),
alpha_factor=a,
lightness_factor=secondary_lightness_factor,
),
secondary_mode,
)
case "frodo640aes":
return cmap(5 / no_algos), primary_mode
case "p256_frodo640aes":
return (
transform_cmap_color(
cmap(5 / no_algos),
alpha_factor=a,
lightness_factor=secondary_lightness_factor,
),
secondary_mode,
)
case "x25519_frodo640aes":
return (
transform_cmap_color(
cmap(5 / no_algos),
alpha_factor=a,
lightness_factor=tertiary_lightness_factor,
),
tertiary_mode,
)
case "frodo640shake":
return cmap(6 / no_algos), primary_mode
case "p256_frodo640shake":
return (
transform_cmap_color(
cmap(6 / no_algos),
alpha_factor=a,
lightness_factor=secondary_lightness_factor,
),
secondary_mode,
)
case "x25519_frodo640shake":
return (
transform_cmap_color(
cmap(6 / no_algos),
alpha_factor=a,
lightness_factor=tertiary_lightness_factor,
),
tertiary_mode,
)
case "frodo976aes":
return cmap(5.3 / no_algos), primary_mode
case "p384_frodo976aes":
return (
transform_cmap_color(
cmap(5.3 / no_algos),
alpha_factor=a,
lightness_factor=secondary_lightness_factor,
),
secondary_mode,
)
case "x448_frodo976aes":
return (
transform_cmap_color(
cmap(5.3 / no_algos),
alpha_factor=a,
lightness_factor=tertiary_lightness_factor,
),
tertiary_mode,
)
case "frodo976shake":
return cmap(6.3 / no_algos), primary_mode
case "p384_frodo976shake":
return (
transform_cmap_color(
cmap(6.3 / no_algos),
alpha_factor=a,
lightness_factor=secondary_lightness_factor,
),
secondary_mode,
)
case "x448_frodo976shake":
return (
transform_cmap_color(
cmap(6.3 / no_algos),
alpha_factor=a,
lightness_factor=tertiary_lightness_factor,
),
tertiary_mode,
)
case "frodo1344aes":
return cmap(5.6 / no_algos), primary_mode
case "p521_frodo1344aes":
return (
transform_cmap_color(
cmap(5.6 / no_algos),
alpha_factor=a,
lightness_factor=secondary_lightness_factor,
),
secondary_mode,
)
case "frodo1344shake":
return cmap(6.6 / no_algos), primary_mode
case "p521_frodo1344shake":
return (
transform_cmap_color(
cmap(6.6 / no_algos),
alpha_factor=a,
lightness_factor=secondary_lightness_factor,
),
secondary_mode,
)
case "x25519_mlkem768":
return (
transform_cmap_color(cmap(7 / no_algos), alpha_factor=a),
primary_mode,
)
case "p256_mlkem768":
return (
transform_cmap_color(
cmap(7.5 / no_algos), lightness_factor=secondary_lightness_factor
),
primary_mode,
)
case "p384_mlkem1024":
return (
transform_cmap_color(
cmap(7.99 / no_algos), lightness_factor=tertiary_lightness_factor
),
primary_mode,
)
case _:
print(f"NO COLOR MATCH FOUND FOR {kem_alg}", file=sys.stderr)
sys.exit(1)
def get_color_and_mode_for_protocol(protocol: str):
primary_mode = "-"
no_protocols = 6
match protocol:
case "quic":
return cmap(0 / no_protocols), primary_mode
case "tlstcp":
return cmap(1 / no_protocols), primary_mode
case "cquiche-reno":
return cmap(2 / no_protocols), primary_mode
case "cquiche-cubic":
return cmap(3 / no_protocols), primary_mode
case "cquiche-bbr":
return cmap(4 / no_protocols), primary_mode
case "cquiche-bbr2":
return cmap(5 / no_protocols), primary_mode
case _:
print(f"NO COLOR MATCH FOUND FOR Protocol {protocol}", file=sys.stderr)
sys.exit(1)
def transform_cmap_color(
color, hue_shift=0, saturation_factor=1, lightness_factor=1, alpha_factor=1
):
def value_between(minimum, value, maximum):
return max(minimum, min(value, maximum))
r, g, b, a = color
h, l, s = colorsys.rgb_to_hls(r, g, b)
h += hue_shift
l *= lightness_factor
l = value_between(0, l, 1)
s *= saturation_factor
s = value_between(0, s, 1)
r, g, b = colorsys.hls_to_rgb(h, l, s)
a *= alpha_factor
return r, g, b, a
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
def get_label_for_scenario(label: str) -> str:
match label:
case "delay":
return "Latenz (ms)"
case "jitter_delay20ms":
return "Jitter (ms)"
case "packetloss":
return "Paketverlustrate (%)"
case "duplicate":
return "Duplikationsrate (%)"
case "corrupt":
return "Korruptierungsrate (%)"
case "reorder":
return "Neuanordnungsrate (%)"
case "rate_both":
return "Bandbreite (Mbit/s)"
case "rate_client":
return "Bandbreite Client (Mbit/s)"
case "rate_server":
return "Bandbreite Server (Mbit/s)"
case "static":
return "Nummerierung der Messung"
case _:
print(f"NO LABEL FOUND FOR {label}", file=sys.stderr)
sys.exit(1)
# plots lines of different statistical values
def plot_lines(data):
def plot_lines_for_sec_level(
data, line_type="median", combined_with_hybrids: bool = False
):
f"{PLOTS_DIR}/lines/per-protocol/{line_type}s-of-sec-level/combined-with-hybrids",
mode=0o777,
exist_ok=True,
)
# get all combination of scenario, protocol, sec_level
unique_combinations = data[
["scenario", "protocol", "sec_level"]
].drop_duplicates()
# print(len(unique_combinations))
# print(unique_combinations)
for _, row in unique_combinations.iterrows():
sec_level = row["sec_level"]
if combined_with_hybrids:
sec_level = map_security_level_hybrid_together(row["sec_level"])
if sec_level is None:
continue
filtered_data = filter_data(
data,
scenario=row["scenario"],
protocol=row["protocol"],
sec_level=sec_level,
)
# print(f"scenario: {row['scenario']}, protocol: {row['protocol']}, sec_level: {row['sec_level']}")
plt.figure()
for idx, kem_alg in enumerate(
filtered_data["kem_alg"].unique().sort_values()
):
# color = cmap(idx / len(filtered_data["kem_alg"].unique()))
color, mode = get_color_and_mode(
kem_alg, combined_with_hybrids=combined_with_hybrids
)
filtered_data_single_kem_alg = filter_data(
filtered_data, kem_alg=kem_alg
)
# print(filtered_data_single_kem_alg)
y = filtered_data_single_kem_alg[line_type]
x = get_x_axis(row["scenario"], filtered_data_single_kem_alg, len(y))
# print(
# f"scenario: {row['scenario']}, protocol: {row['protocol']}, sec_level: {row['sec_level']}, kem_alg: {kem_alg}"
# )
# print(f"x: {x}")
# print(f"y: {y}")
# plt.fill_between(x, filtered_data_single_kem_alg["qtl_25"], filtered_data_single_kem_alg["qtl_75"], alpha=0.2, color=color)
plt.plot(x, y, linestyle=mode, marker=".", color=color, label=kem_alg)
plt.ylim(bottom=0)
plt.xlim(left=0, right=x.max() + (x.max() / 50))
plt.xlabel(get_label_for_scenario(row["scenario"]))
y_label = f"Time-to-first-byte (ms)"
if line_type == "iqr":
y_label = "Interquartilsabstand (ms)"
if line_type == "riqr":
y_label = "Interquartilsabstand / Median"
plt.ylabel(y_label)
# plt.title(
# f"Medians of {row['scenario']} in {row['protocol']} in {row['sec_level']}"
# )
plt.grid()
plt.legend(
bbox_to_anchor=(0.5, 1), loc="lower center", ncol=3, fontsize="small"
)
plt.tight_layout()
subdir = ""
appendix = ""
if combined_with_hybrids:
subdir = "combined-with-hybrids/"
appendix = "-combined-with-hybrids"
plt.savefig(
f"{PLOTS_DIR}/lines/per-protocol/{line_type}s-of-sec-level/{subdir}{line_type}-{row['scenario']}-{row['protocol']}-{row['sec_level']}{appendix}.pdf"
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
def plot_lines_for_tcp_and_cquiche_cubic_for_a_sec_level(data, line_type="median"):
os.makedirs(
f"{PLOTS_DIR}/lines/between-cquiche-cubic-and-tlstcp/{line_type}s-of-sec-level/",
mode=0o777,
exist_ok=True,
)
# get all combination of scenario, protocol, sec_level
unique_combinations = data[["scenario", "sec_level"]].drop_duplicates()
# print(len(unique_combinations))
# print(unique_combinations)
for _, row in unique_combinations.iterrows():
sec_level = row["sec_level"]
filtered_data = filter_data(
data,
scenario=row["scenario"],
sec_level=sec_level,
)
# print(f"scenario: {row['scenario']}, protocol: {row['protocol']}, sec_level: {row['sec_level']}")
plt.figure()
for protocol in ["tlstcp", "cquiche-cubic"]:
inner_filtered_data = filter_data(filtered_data, protocol=protocol)
for idx, kem_alg in enumerate(
inner_filtered_data["kem_alg"].unique().sort_values()
):
# color = cmap(idx / len(inner_filtered_data["kem_alg"].unique()))
color, mode = get_color_and_mode(kem_alg)
if protocol == "tlstcp":
mode = "--"
inner_filtered_data_single_kem_alg = filter_data(
inner_filtered_data, kem_alg=kem_alg
)
# print(inner_filtered_data_single_kem_alg)
y = inner_filtered_data_single_kem_alg[line_type]
x = get_x_axis(
row["scenario"], inner_filtered_data_single_kem_alg, len(y)
)
# print(
# f"scenario: {row['scenario']}, protocol: {row['protocol']}, sec_level: {row['sec_level']}, kem_alg: {kem_alg}"
# )
# print(f"x: {x}")
# print(f"y: {y}")
prefix = "cquiche-cubic-"
if protocol == "tlstcp":
prefix = "tlstcp-"
plt.plot(
x,
y,
linestyle=mode,
marker=".",
color=color,
label=f"{prefix}{kem_alg}",
)
plt.ylim(bottom=0)
plt.xlim(left=0, right=x.max() + (x.max() / 50))
plt.xlabel(get_label_for_scenario(row["scenario"]))
y_label = f"Time-to-first-byte (ms)"
if line_type == "iqr":
y_label = "Interquartilsabstand (ms)"
if line_type == "riqr":
y_label = "Interquartilsabstand / Median"
plt.ylabel(y_label)
# plt.title(
# f"Medians of {row['scenario']} in {row['protocol']} in {row['sec_level']}"
# )
plt.grid()
plt.legend(
bbox_to_anchor=(0.5, 1),
loc="lower center",
ncol=3,
fontsize="small",
)
plt.tight_layout()
plt.savefig(
f"{PLOTS_DIR}/lines/between-cquiche-cubic-and-tlstcp/{line_type}s-of-sec-level/{line_type}-{row['scenario']}-{row['sec_level']}.pdf"
)
plt.close()
def plot_lines_for_comparisons_between_protocols(
data, line_type="median", only_cquiche=False
):
f"{PLOTS_DIR}/lines/between-protocols/comparison-of-{line_type}s/only-cquiche-clients",
mode=0o777,
exist_ok=True,
)
ldata = data
if only_cquiche:
ldata = ldata[ldata["protocol"].str.startswith("cquiche")]
# sec_level is only needed for the filename
unique_combinations = ldata[
["scenario", "kem_alg", "sec_level"]
].drop_duplicates()
for _, row in unique_combinations.iterrows():
filtered_data = filter_data(
scenario=row["scenario"],
kem_alg=row["kem_alg"],
)
plt.figure()
for idx, protocol in enumerate(
filtered_data["protocol"].unique().sort_values()
):
color, mode = get_color_and_mode_for_protocol(protocol)
filtered_data_single_protocol = filter_data(
filtered_data, protocol=protocol
)
y = filtered_data_single_protocol[line_type]
x = get_x_axis(row["scenario"], filtered_data_single_protocol, len(y))
plt.plot(x, y, linestyle=mode, marker=".", color=color, label=protocol)
plt.ylim(bottom=0)
plt.xlim(left=0, right=x.max() + (x.max() / 50))
plt.xlabel(get_label_for_scenario(row["scenario"]))
y_label = f"Time-to-first-byte (ms)"
if line_type == "iqr":
y_label = "Interquartilsabstand (ms)"
if line_type == "riqr":
y_label = "Interquartilsabstand / Median"
plt.ylabel(y_label)
plt.grid()
plt.legend(
bbox_to_anchor=(0.5, 1), loc="lower center", ncol=3, fontsize="small"
)
plt.tight_layout()
subdir = ""
appendix = ""
if only_cquiche:
subdir = "only-cquiche-clients/"
appendix = "-only-cquiche-clients"
f"{PLOTS_DIR}/lines/between-protocols/comparison-of-{line_type}s/{subdir}{line_type}-{row['scenario']}-{row['sec_level']}-{row['kem_alg']}{appendix}.pdf"
)
plt.close()
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
def plot_medians_for_rate_with_cutoff(data, cutoff=5, combined_with_hybrids=False):
print("plot_medians_for_rate")
os.makedirs(
f"{PLOTS_DIR}/lines/medians-rate-with-cutoff/combined-with-hybrids",
mode=0o777,
exist_ok=True,
)
line_type = "median"
for scenario in ["rate_both", "rate_client", "rate_server"]:
# get all combination of scenario, protocol, sec_level
ldata = data
ldata = ldata.query("scenario == @scenario")
unique_combinations = ldata[
["scenario", "protocol", "sec_level"]
].drop_duplicates()
# print(len(unique_combinations))
# print(unique_combinations)
for _, row in unique_combinations.iterrows():
sec_level = row["sec_level"]
if combined_with_hybrids:
sec_level = map_security_level_hybrid_together(row["sec_level"])
if sec_level is None:
continue
filtered_data = filter_data(
data,
scenario=scenario,
protocol=row["protocol"],
sec_level=sec_level,
)
# print(
# f"scenario: {scenario}, protocol: {row['protocol']}, sec_level: {row['sec_level']}"
# )
plt.figure()
for idx, kem_alg in enumerate(
filtered_data["kem_alg"].unique().sort_values()
):
color, mode = get_color_and_mode(
kem_alg, combined_with_hybrids=combined_with_hybrids
)
filtered_data_single_kem_alg = filter_data(
filtered_data, kem_alg=kem_alg
)
y = filtered_data_single_kem_alg[line_type]
x = get_x_axis(scenario, filtered_data_single_kem_alg, len(y))
plt.plot(
x, y, linestyle=mode, marker=".", color=color, label=kem_alg
)
plt.ylim(bottom=0)
plt.xlim(left=0, right=cutoff + 0.2)
plt.xlabel(get_label_for_scenario(scenario))
plt.ylabel("Time-to-first-byte (ms)")
plt.xticks(np.arange(0, cutoff + 0.5, 0.5))
plt.grid()
plt.legend(
bbox_to_anchor=(0.5, 1),
loc="lower center",
ncol=3,
fontsize="small",
)
plt.tight_layout()
subdir = ""
appendix = ""
cutoff_str = f"-cutoff-at-{cutoff}"
if combined_with_hybrids:
subdir = "combined-with-hybrids/"
appendix = "-combined-with-hybrids"
plt.savefig(
f"{PLOTS_DIR}/lines/medians-rate-with-cutoff/{subdir}{line_type}-{scenario}-{row['protocol']}-{row['sec_level']}{cutoff_str}{appendix}.pdf"
)
plt.close()
def plot_qtls_of_single_algorithm(data):
print("plot_qtls_of_single_algorithms")
f"{PLOTS_DIR}/lines/quantiles-of-single-algorithm",
mode=0o777,
exist_ok=True,
# get all combination of scenario, protocol, sec_level, kem_alg
unique_combinations = data[
["scenario", "protocol", "sec_level", "kem_alg"]
].drop_duplicates()
for _, row in unique_combinations.iterrows():
filtered_data = filter_data(
data,
scenario=row["scenario"],
protocol=row["protocol"],
sec_level=row["sec_level"],
kem_alg=row["kem_alg"],
)
# print(f"scenario: {row['scenario']}, protocol: {row['protocol']}, sec_level: {row['sec_level']}, kem_alg: {row['kem_alg']}")
plt.figure()
# plt.fill_between(
# x, filtered_data["qtl_25"], filtered_data["qtl_75"], alpha=0.5
# )
QUANTILES = [1, 5, 10, 25, 40, 50, 60, 75, 90, 95, 99]
color_map = plt.get_cmap("viridis")
measurements = np.array(filtered_data["measurements"])
for qtl in QUANTILES:
color = color_map(1 - (qtl / 100))
y = [np.quantile(x, qtl / 100) for x in measurements]
x = get_x_axis(row["scenario"], filtered_data, len(y))
plt.plot(
x, y, linestyle="-", marker=".", color=color, label=f"{qtl} qtl"
)
plt.ylim(bottom=0)
plt.xlim(left=0, right=x.max() + (x.max() / 50))
plt.xlabel(get_label_for_scenario(row["scenario"]))
plt.ylabel(f"Time-to-first-byte (ms)")
plt.grid()
plt.legend(
bbox_to_anchor=(0.5, 1), loc="lower center", ncol=6, fontsize="small"
)
plt.tight_layout()
# plt.title(
# f"Median of {row['scenario']} in {row['protocol']} in {row['sec_level']} with {row['kem_alg']}"
# )
plt.savefig(
f"{PLOTS_DIR}/lines/quantiles-of-single-algorithm/quantiles-for-{row['scenario']}-{row['protocol']}-{row['sec_level']}-{row['kem_alg']}.pdf"
)
plt.close()
"qtl_25",
"qtl_75",
"qtl_95",
"qtl_99",
"iqr",
"riqr",
"skewness",
"kurtosis",
]
for statistical_measurement in statistical_measurements:
print(f"Generating graphs for {statistical_measurement}")
plot_lines_for_sec_level(
data, line_type=statistical_measurement, combined_with_hybrids=False
)
plot_lines_for_sec_level(
data, line_type=statistical_measurement, combined_with_hybrids=True
)
plot_lines_for_comparisons_between_protocols(
data, line_type=statistical_measurement, only_cquiche=False
)
plot_lines_for_comparisons_between_protocols(
data, line_type=statistical_measurement, only_cquiche=True
plot_lines_for_tcp_and_cquiche_cubic_for_a_sec_level(
data, line_type=statistical_measurement
)
plot_qtls_of_single_algorithm(data)
plot_medians_for_rate_with_cutoff(data, cutoff=5, combined_with_hybrids=False)
plot_medians_for_rate_with_cutoff(data, cutoff=5, combined_with_hybrids=True)
# plots distributions of the individual data points
def plot_distributions(data):
os.makedirs(
f"{PLOTS_DIR}/distributions/single",
mode=0o777,
exist_ok=True,
)
def plot_multiple_violin_plots(data, filtered: bool = False):
os.makedirs(
f"{PLOTS_DIR}/distributions/multiple-violin-plots/filtered/",
mode=0o777,
exist_ok=True,
)
unique_combinations = data[
["scenario", "protocol", "sec_level", "kem_alg"]
].drop_duplicates()
# print(unique_combinations)
for _, row in unique_combinations.iterrows():
filtered_data = filter_data(
data,
scenario=row["scenario"],
protocol=row["protocol"],
sec_level=row["sec_level"],
kem_alg=row["kem_alg"],
)
if row["scenario"] == "static":
continue
# print(
# f"scenario: {row['scenario']}, protocol: {row['protocol']}, sec_level: {row['sec_level']}, kem_alg: {row['kem_alg']}, len: {len(filtered_data)}"
# )
# return the data with removed rows and the ticks for the x-axis
scenario = data.iloc[0]["scenario"]
scenario_column_name = get_x_axis_column_name(scenario)
match scenario:
case "packetloss" | "duplicate" | "reorder" | "corrupt":
# return data where src_packetloss is 0, 4, 8, 12, 16 or 20
ldata = data.query(f"{scenario_column_name} % 4 == 0")
case "jitter_delay20ms":
ldata = data[
data[scenario_column_name].isin([0, 3, 7, 12, 15, 20])
]
case "rate_both" | "rate_client" | "rate_server":
ldata = data[
data[scenario_column_name].isin(
[
0.1,
5,
10,
100,
]
)
]
case "delay":
ldata = data[
data[scenario_column_name].isin([1, 20, 40, 80, 100, 190])
]
print("No case for this scenario:", scenario)
ldata = pd.concat(
[
data.iloc[[0]],
data.iloc[3:-4:4],
data.iloc[[-1]],
]
)
return ldata, ldata[scenario_column_name].to_list()
filtered_data, tick_values = remove_rows_of_data(filtered_data)
# print(filtered_data)
# if filtered_data.iloc[0]["scenario"] == "jitter_delay20ms":
plt.figure()
x = get_x_axis(row["scenario"], filtered_data, len(filtered_data))
x = x.to_list()
# print(x)
vplots = plt.violinplot(
filtered_data["measurements"],
positions=x,
showmedians=False,
showextrema=False,
for pc in vplots["bodies"]:
pc.set_facecolor("blue")
pc.set_edgecolor("darkblue")
# make the median line transparent
# for pc in plt.gca().collections:
# pc.set_alpha(0.5)
plt.ylim(bottom=0)
plt.ylim(bottom=0, top=1.5 * filtered_data["qtl_95"].max())
plt.xticks(tick_values)
plt.xlim(left=-1.5)
plt.xlabel(get_label_for_scenario(row["scenario"]))
plt.ylabel(f"Time-to-first-byte (ms)")
plt.grid()
subdir = "filtered/" if filtered else ""
appendix = "-filtered" if filtered else ""
plt.savefig(
f"{PLOTS_DIR}/distributions/multiple-violin-plots/{subdir}multiple-violin-plots-for-{row['scenario']}-{row['protocol']}-{row['sec_level']}-{row['kem_alg']}{appendix}.pdf"
)
plt.close()
def plot_single_violin_plot(data):
unique_combinations = data[
["scenario", "protocol", "sec_level", "kem_alg"]
].drop_duplicates()
for _, row in unique_combinations.iterrows():
filtered_data = filter_data(
data,
scenario=row["scenario"],
protocol=row["protocol"],
sec_level=row["sec_level"],
kem_alg=row["kem_alg"],
)
# print(
# f"scenario: {row['scenario']}, protocol: {row['protocol']}, sec_level: {row['sec_level']}, kem_alg: {row['kem_alg']}"
# )
for _, row in filtered_data.iterrows():
if row["scenario"] == "static":
continue
value = get_x_axis(row["scenario"], row, 1)
# print(value)
plt.figure()
plt.violinplot(row["measurements"], showmedians=True)
# plt.ylim(bottom=0)
# plt.xlim(left=0, right=x.max() + (x.max() / 50))
plt.xlabel("Dichte")
plt.ylabel(f"Time-to-first-byte (ms)")
plt.grid()
plt.savefig(
f"{PLOTS_DIR}/distributions/single/single-violin-plot-for-{row['scenario']}-{row['protocol']}-{row['sec_level']}-{row['kem_alg']}-{value}.pdf"
)
plt.close()
# return
Bartolomeo Berend Müller
committed
def plot_cdf_of_sec_level(data, cutoff=None):
subdir_string = ""
graph_name_extension = ""
if cutoff is not None:
subdir_string = f"cutoff-at-{cutoff}/"
graph_name_extension = f"-cutoff-at-{cutoff}"
Bartolomeo Berend Müller
committed
Bartolomeo Berend Müller
committed
f"{PLOTS_DIR}/distributions/cdf/per-sec-level/{subdir_string}",
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
mode=0o777,
exist_ok=True,
)
unique_combinations = data[
["scenario", "protocol", "sec_level"]
].drop_duplicates()
# filter out every scenario that is not packetloss or corrupt
unique_combinations = unique_combinations[
unique_combinations["scenario"].isin(["packetloss", "corrupt"])
]
# print(unique_combinations)
for _, row in unique_combinations.iterrows():
filtered_data = filter_data(
data,
scenario=row["scenario"],
protocol=row["protocol"],
sec_level=row["sec_level"],
)
# print(filtered_data)
def get_row_name_for_scenario(scenario):
match scenario:
case "packetloss":
return "srv_pkt_loss"
case "corrupt":
return "srv_corrupt"
case _:
print("No case for this scenario:", row["scenario"])
exit(1)
row_name = get_row_name_for_scenario(row["scenario"])
scenario_param_values = filtered_data[row_name].unique()
# print(scenario_param_values)
for param_value in scenario_param_values:
filtered_data_single_param_value = filtered_data.query(
f"{row_name} == {param_value}"
)
# print(filtered_data_single_param_value)
plt.figure()
for _, row in filtered_data_single_param_value.iterrows():
color, mode = get_color_and_mode(
row["kem_alg"], combined_with_hybrids=False
)
plt.ecdf(
row["measurements"],
label=row["kem_alg"],
color=color,
linestyle=mode,
)
# plt.ylim(bottom=0)
Bartolomeo Berend Müller
committed
if cutoff is not None:
plt.xlim(0, cutoff)
# plt.xlim(left=0, right=x.max() + (x.max() / 50))
plt.ylabel("Wahrscheinlichkeit")
plt.xlabel(f"Time-to-first-byte (ms)")
plt.grid()
plt.legend(
bbox_to_anchor=(0.5, 1),
loc="lower center",
ncol=3,
fontsize="small",
)
plt.tight_layout()
Bartolomeo Berend Müller
committed
f"{PLOTS_DIR}/distributions/cdf/per-sec-level/{subdir_string}cdf-for-{row['scenario']}-{row['protocol']}-{row['sec_level']}-{param_value}{graph_name_extension}.pdf"
)
plt.close()
plot_multiple_violin_plots(data, filtered=False)
plot_multiple_violin_plots(data, filtered=True)
# plot_single_violin_plot(data) # takes an age
Bartolomeo Berend Müller
committed
plot_cdf_of_sec_level(data, cutoff=None)
plot_cdf_of_sec_level(data, cutoff=2000)
plot_cdf_of_sec_level(data, cutoff=5000)
# TODO make a violinplot/eventplot for many algos in static scenario
def plot_static_data(data):
os.makedirs(f"{PLOTS_DIR}/static/single", mode=0o777, exist_ok=True)
def plot_static_data_for_multiple_algorithms(data):
unique_combinations = data[["protocol", "sec_level"]].drop_duplicates()
for idx, row in unique_combinations.iterrows():
sec_level = map_security_level_hybrid_together(row["sec_level"])
if sec_level is None:
continue
filtered_data = filter_data(
data,
scenario="static",
protocol=row["protocol"],
sec_level=sec_level,
)
iqrs = pd.DataFrame()
# plt.figure(figsize=(6, 6))
plt.figure()
kem_algs = []
for idx, kem_alg in enumerate(
filtered_data["kem_alg"].unique().sort_values(),
):
kem_algs.append(kem_alg)
filtered_data_single_kem_alg = filter_data(
filtered_data, kem_alg=kem_alg
)
plt.boxplot(
filtered_data_single_kem_alg["median"],
positions=[idx],
widths=0.6,
)
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
iqrs = pd.concat(
[
iqrs,
pd.DataFrame(
{
"kem_alg": [kem_alg],
"iqr": [
scipy.stats.iqr(
filtered_data_single_kem_alg["median"]
)
],
}
),
],
ignore_index=True,
)
# print(
# f"IQR for {kem_alg}: {scipy.stats.iqr(filtered_data_single_kem_alg['median'])}"
# )
# Get the median irqs for all algorithms
# print(row["protocol"], row["sec_level"])
# print(iqrs)
# print(iqrs["iqr"].describe())
# print("Median:", iqrs["iqr"].median())
# print("IQR:", scipy.stats.iqr(iqrs["iqr"]))
# print()
plt.xticks(
range(len(filtered_data["kem_alg"].unique())),
kem_algs,
rotation=45,
ha="right",
)
plt.xlabel("KEM Algorithms")
plt.ylabel("Time-to-first-byte (ms)")
plt.grid()
sec_level_string = (
sec_level if type(sec_level) == str else "-".join(sec_level)
)
plt.tight_layout()
plt.savefig(
os.path.join(
PLOTS_DIR,
"static",
f"boxplots-of-medians-for-static-{row['protocol']}-{sec_level_string}.pdf",
)
)
plt.close()
def plot_static_data_for_single_algorithms(data):
unique_combinations = data[
["scenario", "protocol", "sec_level", "kem_alg"]
].drop_duplicates()
unique_combinations = filter_data(unique_combinations, scenario="static")
for idx, row in unique_combinations.iterrows():
filtered_data = filter_data(
data,
scenario="static",
protocol=row["protocol"],
sec_level=row["sec_level"],
kem_alg=row["kem_alg"],
)
def boxplot_of_medians_for_configuration(filtered_data, row):
plt.figure()
plt.boxplot(filtered_data["median"])
plt.grid()
plt.savefig(
os.path.join(
PLOTS_DIR,
"static",
"single",
f"boxplot-of-medians-for-{row['scenario']}-{row['protocol']}-{row['sec_level']}-{row['kem_alg']}.pdf",
)
plt.close()
# why the density of violin plot and kde plot differ, while using the same scott kde just sideways:
# Dove deep into the implementation from matplotlib and scipy, and they seem to calculate scotts factor in the same way, so dunno
def condensed_violin_plot_for_configuration(filtered_data, row):
# for multiple runs of the same static scenario, data taken together
measurements_flattend = filtered_data["measurements"].explode().tolist()
# print(filtered_data["measurements"].explode())
# print(len(measurements_flattend))
plt.figure()
plt.violinplot(measurements_flattend, showmedians=True)
plt.ylabel("Time-to-first-byte (ms)")
plt.xlabel("Dichte")
plt.grid()
plt.savefig(
os.path.join(
PLOTS_DIR,
"static",
"single",
f"condensed-violin-plot-for-{len(measurements_flattend)}-measurements-of-{row['scenario']}-{row['protocol']}-{row['sec_level']}-{row['kem_alg']}.pdf",
)
plt.close()
def condensed_histogram_plot_for_configuration(filtered_data, row):
measurements_flattend = filtered_data["measurements"].explode().tolist()
plt.figure()
plt.hist(measurements_flattend, bins=100, density=True)
plt.xlabel("Time-to-first-byte (ms)")
plt.ylabel("Dichte")
plt.grid()
plt.savefig(
os.path.join(
PLOTS_DIR,
"static",
"single",
f"condensed-histogram-plot-for-{len(measurements_flattend)}-measurements-of-{row['scenario']}-{row['protocol']}-{row['sec_level']}-{row['kem_alg']}.pdf",
)
)
plt.close()
def condensed_kernel_density_estimate_plot_for_configuration(
filtered_data, row
):
measurements_flattend = filtered_data["measurements"].explode().tolist()
plt.figure()
kde = scipy.stats.gaussian_kde(measurements_flattend)
xmin = min(measurements_flattend) - 0.2
xmax = max(measurements_flattend) + 0.1
x = np.linspace(
xmin,
xmax,
1000,
kde_values = kde(x)
plt.plot(x, kde_values)
plt.fill_between(x, kde_values, alpha=0.5)
plt.xlabel("Time-to-first-byte (ms)")
plt.ylabel("Dichte")
plt.xlim([xmin, xmax])
plt.ylim([0, max(kde_values) + 0.1])
plt.grid()
plt.savefig(
os.path.join(
PLOTS_DIR,
"static",
"single",
f"condensed-kde-plot-for-{len(measurements_flattend)}-measurements-of-{row['scenario']}-{row['protocol']}-{row['sec_level']}-{row['kem_alg']}.pdf",
)
)
plt.close()
boxplot_of_medians_for_configuration(filtered_data, row)
condensed_violin_plot_for_configuration(filtered_data, row)
condensed_histogram_plot_for_configuration(filtered_data, row)
condensed_kernel_density_estimate_plot_for_configuration(filtered_data, row)
# return
plot_static_data_for_multiple_algorithms(data)
plot_static_data_for_single_algorithms(data)
def plot_general_plots():
def get_color_for_kem_algo(kem_algo):
if "mlkem" in kem_algo:
return "blue"
if "bikel" in kem_algo:
return "red"
if "hqc" in kem_algo:
return "green"
if "frodo" in kem_algo:
return "orange"
return "grey"
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
def plot_oqs_performance():
def plot_liboqs_performance(liboqs_speed_data):
os.makedirs(
"plots/general/oqs-performance/liboqs", mode=0o777, exist_ok=True
)
# print(liboqs_speed_data.columns)
# print(
# liboqs_speed_data[
# [
# "algorithm",
# "keygen_mean_time_us",
# "encaps_mean_time_us",
# "decaps_mean_time_us",
# "total_liboqs_performance_time_us",
# ]
# ]
# )
# print(
# liboqs_speed_data[
# [
# "algorithm",
# "total_liboqs_performance_cpu_cycles",
# "total_expected_cycles",
# "total_liboqs_performance_ratio",
# ]
# ]
# )
# figure with all algorithms with times for each operation
plt.figure()
fig, ax = plt.subplots()
bar_width = 0.5
index = np.arange(len(liboqs_speed_data))
color_map = plt.get_cmap("viridis")
bar1 = ax.bar(
index,
liboqs_speed_data["keygen_mean_time_us"],
bar_width,
label="Keygen",
color=color_map(0.1),
)
bar2 = ax.bar(
index,
liboqs_speed_data["encaps_mean_time_us"],
bar_width,
bottom=liboqs_speed_data["keygen_mean_time_us"],
label="Encaps",
color=color_map(0.4),
)
bar3 = ax.bar(
index,
liboqs_speed_data["decaps_mean_time_us"],
bar_width,
bottom=liboqs_speed_data["keygen_mean_time_us"]
+ liboqs_speed_data["encaps_mean_time_us"],
label="Decaps",
color=color_map(0.7),
)
values = liboqs_speed_data["total_liboqs_performance_time_us"]
for idx, rect in enumerate(bar3):
ax.text(
rect.get_x() + rect.get_width() / 2.0,
values.iloc[idx],
f"{values.iloc[idx]:.1f}",
ha="center",
va="bottom",
fontsize=8,
)
# ax.set_xlabel("Algorithm")
ax.set_ylabel("Time (µs)")
# ax.set_title("Keygen, Encaps, and Decaps Time by Algorithm")
ax.set_xticks(index)
ax.set_xticklabels(liboqs_speed_data["algorithm"], rotation=60, ha="right")
ax.legend()
ax.grid()
ax.set_axisbelow(True)
ax.set_ylim(
0, liboqs_speed_data["total_liboqs_performance_time_us"].max() * 1.1
)
plt.tight_layout()
plt.savefig(
"plots/general/oqs-performance/liboqs/time-for-operations-for-each-algo.pdf"
)
plt.close()
# figure with all algorithms with cycles for keygen, encaps and decaps, next to it the expected cycles for each operation
liboqs_speed_data_bikel5_changed = liboqs_speed_data.copy()
liboqs_speed_data_bikel5_changed.loc[
liboqs_speed_data_bikel5_changed["algorithm"] == "BIKE-L5",
"keygen_expected_cycles",
] = 0
liboqs_speed_data_bikel5_changed.loc[
liboqs_speed_data_bikel5_changed["algorithm"] == "BIKE-L5",
"encaps_expected_cycles",
] = 0
liboqs_speed_data_bikel5_changed.loc[
liboqs_speed_data_bikel5_changed["algorithm"] == "BIKE-L5",
"decaps_expected_cycles",
] = 0
plt.figure()
fig, ax = plt.subplots()
bar_width = 0.35
index = np.arange(len(liboqs_speed_data_bikel5_changed))
# bar1 is for measured cycles, bar2 is for expected cycles
bar11 = ax.bar(
index - 0.01 - bar_width / 2,
liboqs_speed_data_bikel5_changed["keygen_mean_cpu_cycles"],
bar_width,
label="Keygen",
color=color_map(0.1),
)
bar12 = ax.bar(
index - 0.01 - bar_width / 2,
liboqs_speed_data_bikel5_changed["encaps_mean_cpu_cycles"],
bar_width,
bottom=liboqs_speed_data_bikel5_changed["keygen_mean_cpu_cycles"],
label="Encaps",
color=color_map(0.4),
)
bar13 = ax.bar(
index - 0.01 - bar_width / 2,
liboqs_speed_data_bikel5_changed["decaps_mean_cpu_cycles"],
bar_width,
bottom=liboqs_speed_data_bikel5_changed["keygen_mean_cpu_cycles"]
+ liboqs_speed_data_bikel5_changed["encaps_mean_cpu_cycles"],
label="Decaps",
color=color_map(0.7),
)
bar21 = ax.bar(
index + 0.01 + bar_width / 2,
liboqs_speed_data_bikel5_changed["keygen_expected_cycles"],
bar_width,
color=color_map(0.1),
)
bar22 = ax.bar(
index + 0.01 + bar_width / 2,
liboqs_speed_data_bikel5_changed["encaps_expected_cycles"],
bar_width,
bottom=liboqs_speed_data_bikel5_changed["keygen_expected_cycles"],
color=color_map(0.4),
)
bar23 = ax.bar(
index + 0.01 + bar_width / 2,
liboqs_speed_data_bikel5_changed["decaps_expected_cycles"],
bar_width,
bottom=liboqs_speed_data_bikel5_changed["keygen_expected_cycles"]
+ liboqs_speed_data_bikel5_changed["encaps_expected_cycles"],
color=color_map(0.7),
)
values = liboqs_speed_data_bikel5_changed[
"total_liboqs_performance_cpu_cycles"
]
for idx, rect in enumerate(bar13):
ax.text(
rect.get_x() + rect.get_width() / 2.0,
values.iloc[idx] + 100000,
f"{int(values.iloc[idx])}",
ha="center",
va="bottom",
fontsize=8,
rotation=90,
)
values = liboqs_speed_data_bikel5_changed["total_expected_cycles"]
for idx, rect in enumerate(bar23):
if "BIKE-L5" == liboqs_speed_data_bikel5_changed["algorithm"].iloc[idx]:
continue
ax.text(
rect.get_x() + rect.get_width() / 2.0,
values.iloc[idx] + 100000,
f"{int(values.iloc[idx])}",
ha="center",
va="bottom",
fontsize=8,
rotation=90,
)
ax.set_ylabel("CPU Zyklen in Millionen")
ax.yaxis.set_major_formatter(
ticker.FuncFormatter(lambda x, _: f"{int(x/1_000_000)}")
)
ax.set_xticks(index)
ax.set_xticklabels(
liboqs_speed_data_bikel5_changed["algorithm"], rotation=60, ha="right"
)
ax.legend()
ax.grid()
ax.set_axisbelow(True)
ax.set_ylim(
0,
max(
liboqs_speed_data_bikel5_changed[
"total_liboqs_performance_cpu_cycles"
].max(),
liboqs_speed_data_bikel5_changed["total_expected_cycles"].max(),
)
* 1.3,
)
plt.tight_layout()
plt.savefig(
"plots/general/oqs-performance/liboqs/cycles-for-operations-for-each-algo-next-to-expected-cycles.pdf"
)
plt.close()
def plot_openssl_performance(openssl_speed_data, secLevel=""):
os.makedirs(
"plots/general/oqs-performance/openssl", mode=0o777, exist_ok=True
)
# print(openssl_speed_data.columns)
color_map = plt.get_cmap("inferno")
plt.figure()
fig, ax = plt.subplots()
bar_width = 0.35
index = np.arange(len(openssl_speed_data))
# bar1 is for measured cycles, bar2 is for expected cycles
bar11 = ax.bar(
index - 0.01 - bar_width / 2,
openssl_speed_data["keygen_mean_time_s"],
bar_width,
label="Keygen",
color=color_map(0.1),
)
bar12 = ax.bar(
index - 0.01 - bar_width / 2,
openssl_speed_data["encaps_mean_time_s"],
bar_width,
bottom=openssl_speed_data["keygen_mean_time_s"],
label="Encaps",
color=color_map(0.4),
)
bar13 = ax.bar(
index - 0.01 - bar_width / 2,
openssl_speed_data["decaps_mean_time_s"],
bar_width,
bottom=openssl_speed_data["keygen_mean_time_s"]
+ openssl_speed_data["encaps_mean_time_s"],
label="Decaps",
color=color_map(0.7),
)
bar21 = ax.bar(
index + 0.01 + bar_width / 2,
openssl_speed_data["keygen_mean_time_s_expected"],
bar_width,
color=color_map(0.1),
)
bar22 = ax.bar(
index + 0.01 + bar_width / 2,
openssl_speed_data["encaps_mean_time_s_expected"],
bar_width,
bottom=openssl_speed_data["keygen_mean_time_s_expected"],
color=color_map(0.4),
)
bar23 = ax.bar(
index + 0.01 + bar_width / 2,
openssl_speed_data["decaps_mean_time_s_expected"],
bar_width,
bottom=openssl_speed_data["keygen_mean_time_s_expected"]
+ openssl_speed_data["encaps_mean_time_s_expected"],
color=color_map(0.7),
)
values = openssl_speed_data["total_actual_performance_time_s"]
for idx, rect in enumerate(bar13):
ax.text(
rect.get_x() + rect.get_width() / 2.0,
values.iloc[idx] + 0.0001,
f"{values.iloc[idx]:.6f}",
ha="center",
va="bottom",
fontsize=8,
rotation=90,
)
values = openssl_speed_data["total_expected_performance_time_s"]
for idx, rect in enumerate(bar23):
ax.text(
rect.get_x() + rect.get_width() / 2.0,
values.iloc[idx] + 0.0001,
f"{values.iloc[idx]:.6f}",
ha="center",
va="bottom",
fontsize=8,
rotation=90,
)
ax.set_ylabel("Zeit in Sekunden")
ax.set_xticks(index)
ax.set_xticklabels(openssl_speed_data["algorithm"], rotation=60, ha="right")
ax.legend()
ax.grid()
ax.set_axisbelow(True)
ax.set_ylim(
0,
max(
openssl_speed_data["total_actual_performance_time_s"].max(),
openssl_speed_data["total_expected_performance_time_s"].max(),
)
* 1.3,
)
plt.tight_layout()
plt.savefig(
f"plots/general/oqs-performance/openssl/time-vs-expected-time{secLevel}.pdf"
)
plt.close()
liboqs_speed_data = pd.read_feather(f"saved/feathers/liboqs_speed_vm.feather")
liboqs_speed_data = (
performance_eval_of_oqs.concat_liboqs_speed_data_with_expected_values(
liboqs_speed_data
)
)
plot_liboqs_performance(liboqs_speed_data)
openssl_speed_data = pd.read_feather(f"saved/feathers/openssl_speed_vm.feather")
openssl_speed_data = performance_eval_of_oqs.analyze_openssl_speed_data(
openssl_speed_data
)
openssl_speed_data = openssl_speed_data.copy()
openssl_speed_data = openssl_speed_data[
openssl_speed_data["algorithm"].str.contains("_")
]
plot_openssl_performance(openssl_speed_data)
openssl_speed_data_secLevel1 = openssl_speed_data.copy()
openssl_speed_data_secLevel1 = openssl_speed_data_secLevel1[
openssl_speed_data_secLevel1["algorithm"].str.contains("p256_|x25519_")
]
plot_openssl_performance(openssl_speed_data_secLevel1, secLevel="-1")
openssl_speed_data_secLevel35 = openssl_speed_data.copy()
openssl_speed_data_secLevel35 = openssl_speed_data_secLevel35[
openssl_speed_data_secLevel35["algorithm"].str.contains("p384_|p521_|x448_")
]
plot_openssl_performance(openssl_speed_data_secLevel35, secLevel="-35")
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
def plot_send_bytes_against_kem_performance(df, with_hybrids: bool):
if not with_hybrids:
# filter out all hybrids, otherwise the plot is too cluttered
df = df[~df["kem_algo"].str.contains("_")]
plt.figure()
# plt.scatter(df["bytes_sent"], df["performance_us"])
for kem_algo in df["kem_algo"]:
subset = df[df["kem_algo"] == kem_algo]
color = get_color_for_kem_algo(kem_algo)
plt.scatter(
subset["bytes_sent"],
subset["performance_us"],
color=color,
label=kem_algo,
alpha=0.7,
)
for i, txt in enumerate(df["kem_algo"]):
plt.annotate(
txt,
(df["bytes_sent"].iloc[i], df["performance_us"].iloc[i]),
xytext=(0, -3),
textcoords="offset points",
fontsize=8,
ha="center",
va="bottom",
)
plt.xscale("log")
plt.yscale("log")
plt.xlim(30)
plt.ylim(30)
# custom tick stuff from claude
def custom_ticks(start, end):
ticks = [start] + [
10**i for i in range(int(np.log10(start)) + 1, int(np.log10(end)) + 1)
]
return ticks
x_ticks = custom_ticks(30, df["bytes_sent"].max())
y_ticks = custom_ticks(30, df["performance_us"].max())
plt.xticks(x_ticks, [f"{int(x):,}" for x in x_ticks])
plt.yticks(y_ticks, [f"{int(y):,}" for y in y_ticks])
plt.xlabel("Übertragungslänge (Bytes)")
plt.ylabel("Performanz der Operationen (µs)")
plt.grid()
"scatter-of-bytes-sent-against-kem-performance-with-hybrids.pdf"
else "scatter-of-bytes-sent-against-kem-performance.pdf"
)
plt.savefig(
os.path.join(PLOTS_DIR, "general", name),
dpi=300,
)
plt.close()
# print(df)
def plot_public_key_length_against_ciphertext_length(
df, with_hybrids: bool, with_lines: bool
):
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
if not with_hybrids:
# filter out all hybrids, otherwise the plot is too cluttered
df = df[~df["kem_algo"].str.contains("_")]
# only keep one for frodo, since they are the same and remove the hash algo
df = df[~df["kem_algo"].str.contains("shake")]
df.loc[:, "kem_algo"] = df["kem_algo"].apply(lambda x: x.replace("aes", ""))
plt.figure()
for kem_algo in df["kem_algo"]:
subset = df[df["kem_algo"] == kem_algo]
color = get_color_for_kem_algo(kem_algo)
plt.scatter(
subset["length_public_key"],
subset["length_ciphertext"],
color=color,
label=kem_algo,
alpha=0.7,
)
for i, txt in enumerate(df["kem_algo"]):
annotate_offset = (0, -3)
if "bikel1" in txt:
annotate_offset = (0, 0)
if "mlkem1024" in txt:
annotate_offset = (0, -7)
plt.annotate(
txt,
(df["length_public_key"].iloc[i], df["length_ciphertext"].iloc[i]),
xytext=annotate_offset,
textcoords="offset points",
fontsize=8,
ha="center",
va="bottom",
)
if with_lines:
# reason for these magic numbers in obsidian note [[Packet lengths in QUIC over Ethernet]]
plt.axvline(x=940, color="purple", linestyle="--", label="1 Paket Grenze")
plt.axhline(y=277, color="purple", linestyle="--", label="1 Paket Grenze")
plt.axvline(
x=940 + 1157, color="purple", linestyle="--", label="2 Paket Grenze"
)
plt.axhline(
y=277 + 1100, color="purple", linestyle="--", label="2 Paket Grenze"
)
plt.xscale("log")
plt.yscale("log")
plt.xlim(10)
plt.ylim(10)
plt.xlabel("Public Key Länge in Bytes")
plt.ylabel("Ciphertext Länge in Bytes")
plt.grid()
plt.gca().xaxis.set_major_formatter(ticker.ScalarFormatter())
plt.gca().yaxis.set_major_formatter(ticker.ScalarFormatter())
with_hybrids_string = "-with-hybrids" if with_hybrids else ""
with_lines_string = "-with-lines" if with_lines else ""
name = f"scatter-of-public-key-against-ciphertext-length{with_hybrids_string}{with_lines_string}.pdf"
plt.savefig(
os.path.join(PLOTS_DIR, "general", name),
dpi=300,
)
plt.close()
# This does not yet seem like a good idea
# TODO use the riqr to make some graphs
def plot_median_against_iqr(data):
plt.figure()
plt.hexbin(data["median"], data["iqr"], gridsize=50)
print(data["iqr"].describe())
print(data["median"].describe())
# get the line with the maximum median
max_median = data["median"].idxmax()
print(data.iloc[max_median])
plt.grid()
plt.savefig(f"{PLOTS_DIR}/general/median_against_iqr_hexbin.pdf")
plt.close()
plt.figure()
plt.hist2d(data["median"], data["iqr"], bins=50)
plt.grid()
plt.savefig(f"{PLOTS_DIR}/general/median_against_iqr_hist2d.pdf")
plt.close()
os.makedirs(f"{PLOTS_DIR}/general", mode=0o777, exist_ok=True)
df = helper_functions.prepare_kem_performance_data_for_plotting(
helper_functions.get_kem_performance_data()
)
plot_oqs_performance()
plot_send_bytes_against_kem_performance(df, with_hybrids=False)
plot_send_bytes_against_kem_performance(df, with_hybrids=True)
plot_public_key_length_against_ciphertext_length(
df, with_hybrids=False, with_lines=True
)
plot_public_key_length_against_ciphertext_length(
df, with_hybrids=False, with_lines=False
)
plot_public_key_length_against_ciphertext_length(
df, with_hybrids=True, with_lines=True
)
plot_public_key_length_against_ciphertext_length(
df, with_hybrids=True, with_lines=False
)
Loading
Loading full blame...