FuzzBench: collisions report

experiment summary

We show two different aggregate (cross-benchmark) rankings of fuzzers. The first is based on the average of per-benchmarks scores, where the score represents the percentage of the highest reached median bug-coverage on a given benchmark (higher value is better). The second ranking shows the average rank of fuzzers, after we rank them on each benchmark according to their median reached bug-covereges (lower value is better).

By avg. score

	average normalized score
fuzzer
afl_collision_free	96.52
afl	89.44

By avg. rank

	average rank
fuzzer
afl_collision_free	1.44
afl	1.56

Critical difference diagram

The diagram visualizes the average rank of fuzzers (second ranking above) while showing the significance of the differences as well. What is considered a "critical difference" (CD) is based on the Friedman/Nemenyi post-hoc test. See more in the documentation.

Note: If a fuzzer does not support all benchmarks, its ranking as shown in this diagram can be lower than it should be. So please check the list of supported benchmarks for the fuzzer(s) of your interest. The list could be specified in the fuzzer's README.md like this.

Median relative code-coverages on each benchmark

Note: The relative coverage summary table shows the median relative performance of each fuzzer to the experiment maximum. Thus the highest relative performance may not be 100%.
trial_relative_coverage = trial_coverage / experiment_max_coverage

	afl_collision_free	afl
FuzzerMedian	96.95	97.15
FuzzerMean	95.64	95.25
arrow_parquet-arrow-fuzz	97.53	97.31
aspell_aspell_fuzzer	99.80	99.75
ffmpeg_ffmpeg_demuxer_fuzzer	nan	nan
file_magic_fuzzer	98.82	96.96
grok_grk_decompress_fuzzer	93.20	92.97
libarchive_libarchive_fuzzer	90.51	86.37
libgit2_objects_fuzzer	99.80	99.75
libhevc_hevc_dec_fuzzer	80.91	81.13
libhtp_fuzz_htp	99.97	99.91
libxml2_libxml2_xml_reader_for_file_fuzzer	93.01	89.91
matio_matio_fuzzer	97.30	97.80
mruby-2018-05-23	93.65	93.12
muparser_set_eval_fuzzer	97.70	97.70
njs_njs_process_script_fuzzer	96.60	97.19
openh264_decoder_fuzzer	99.81	99.65
php_php-fuzz-execute	96.22	94.27
php_php-fuzz-parser-2020-07-25	99.64	99.47
poppler_pdf_fuzzer	98.85	98.80
proj4_standard_fuzzer	100.00	100.00
quickjs_eval-2020-01-05	93.14	95.23
stb_stbi_read_fuzzer	92.39	91.21
systemd_fuzz-varlink	88.48	88.48
usrsctp_fuzzer_connect	94.71	94.16
wireshark_fuzzshark_ip	95.86	97.11
zstd_stream_decompress	97.43	97.74

Fuzzers are sorted by "FuzzerMean" (average median relative coverage), highest on the left.
Green background = highest relative median coverage.
Blue gradient background = greater than 95% relative median coverage.

Median relative bug-coverages on each benchmark

Note: The relative coverage summary table shows the median relative performance of each fuzzer to the experiment maximum. Thus the highest relative performance may not be 100%.
trial_relative_coverage = trial_coverage / experiment_max_coverage

	afl_collision_free	afl
FuzzerMedian	60.00	50.00
FuzzerMean	56.40	52.62
arrow_parquet-arrow-fuzz	93.75	85.42
aspell_aspell_fuzzer	100.00	100.00
ffmpeg_ffmpeg_demuxer_fuzzer	60.00	70.00
file_magic_fuzzer	0.00	0.00
grok_grk_decompress_fuzzer	14.29	14.29
libarchive_libarchive_fuzzer	nan	nan
libgit2_objects_fuzzer	66.67	33.33
libhevc_hevc_dec_fuzzer	85.94	87.50
libhtp_fuzz_htp	60.00	80.00
libxml2_libxml2_xml_reader_for_file_fuzzer	15.38	15.38
matio_matio_fuzzer	83.33	83.33
mruby-2018-05-23	60.00	40.00
muparser_set_eval_fuzzer	nan	nan
njs_njs_process_script_fuzzer	66.67	33.33
openh264_decoder_fuzzer	66.67	50.00
php_php-fuzz-execute	50.00	50.00
php_php-fuzz-parser-2020-07-25	50.00	37.50
poppler_pdf_fuzzer	82.35	82.35
proj4_standard_fuzzer	nan	nan
quickjs_eval-2020-01-05	50.00	50.00
stb_stbi_read_fuzzer	85.71	78.57
systemd_fuzz-varlink	100.00	100.00
usrsctp_fuzzer_connect	0.00	0.00
wireshark_fuzzshark_ip	50.00	66.67
zstd_stream_decompress	0.00	0.00

Fuzzers are sorted by "FuzzerMean" (average median relative coverage), highest on the left.
Green background = highest relative median coverage.
Blue gradient background = greater than 95% relative median coverage.

Total unique bugs found on each benchmark

	Total	afl	afl_collision_free
FuzzerSum	311	277	276
arrow_parquet-arrow-fuzz	74	65	70
aspell_aspell_fuzzer	2	2	2
ffmpeg_ffmpeg_demuxer_fuzzer	32	25	23
file_magic_fuzzer	1	1	1
grok_grk_decompress_fuzzer	8	8	7
libarchive_libarchive_fuzzer	0	0	0
libgit2_objects_fuzzer	3	3	3
libhevc_hevc_dec_fuzzer	34	34	34
libhtp_fuzz_htp	5	5	5
libxml2_libxml2_xml_reader_for_file_fuzzer	18	13	18
matio_matio_fuzzer	17	15	15
mruby-2018-05-23	9	8	8
muparser_set_eval_fuzzer	0	0	0
njs_njs_process_script_fuzzer	5	3	5
openh264_decoder_fuzzer	6	5	6
php_php-fuzz-execute	2	1	2
php_php-fuzz-parser-2020-07-25	10	10	8
poppler_pdf_fuzzer	46	41	35
proj4_standard_fuzzer	0	0	0
quickjs_eval-2020-01-05	2	1	2
stb_stbi_read_fuzzer	25	25	24
systemd_fuzz-varlink	1	1	1
usrsctp_fuzzer_connect	1	1	1
wireshark_fuzzshark_ip	9	9	5
zstd_stream_decompress	1	1	1

Fuzzers are sorted by "FuzzerSum", highest on the left.
Green background = most unique bugs found.
*note: This table represents unique bugs found across all trials.

arrow_parquet-arrow-fuzz summary

Discovered bug coverage distribution

Mean code coverage growth over time

Mean bug coverage growth over time

* The error bands show the 95% confidence interval around the mean code coverage.

Sample statistics and statistical significance (bugs covered)

Bug coverage sample statistics

		count	mean	std	min	25%	median	75%	max
fuzzer	time
afl_collision_free	82800	14.0	43.214286	8.001717	16.0	44.0	45.0	46.0	48.0
afl	82800	11.0	38.454545	9.709414	10.0	40.0	41.0	42.5	44.0

Vargha-Delaney A12 measure

The table summarizes the A12 values from the pairwise Vargha-Delaney A measure of effect size. Green cells indicate the probability the fuzzer in the row will outperform the fuzzer in the column.

Mann-Whitney U test

The table summarizes the p values of pairwise Mann-Whitney U tests. Green cells indicate that the reached coverage distribution of a given fuzzer pair is significantly different.

Sample statistics and statistical significance (code coverage)

Code coverage sample statistics

		count	mean	std	min	25%	median	75%	max
fuzzer	time
afl_collision_free	82800	14.0	5107.500000	49.310242	5032.0	5075.0	5104.0	5129.25	5214.0
afl	82800	11.0	5099.272727	56.505028	5022.0	5063.0	5092.0	5120.50	5233.0

Vargha-Delaney A12 measure

The table summarizes the A12 values from the pairwise Vargha-Delaney A measure of effect size. Green cells indicate the probability the fuzzer in the row will outperform the fuzzer in the column.

Mann-Whitney U test

The table summarizes the p values of pairwise Mann-Whitney U tests. Green cells indicate that the reached coverage distribution of a given fuzzer pair is significantly different.