FuzzBench: jan13-lab 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 code-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 code-covereges (lower value is better).
By avg. score
average normalized score
fuzzer
bandfuzz_syncmore 99.99
bandfuzz_syncless 99.28
aflplusplus 96.25
By avg. rank
average rank
fuzzer
bandfuzz_syncmore 1.17
aflplusplus 2.00
bandfuzz_syncless 2.67
  • 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

      bandfuzz_syncmore bandfuzz_syncless aflplusplus
    FuzzerMedian 97.50 96.50 97.00
    FuzzerMean 97.50 96.50 93.83
    bloaty_fuzz_target 97.00 94.00 96.00
    curl_curl_fuzzer_http 98.00 98.00 98.00
    freetype2_ftfuzzer 96.00 95.00 82.00
    harfbuzz_hb-shape-fuzzer 99.00 98.00 99.00
    lcms_cms_transform_fuzzer 96.00 95.00 89.00
    libjpeg-turbo_libjpeg_turbo_fuzzer 99.00 99.00 99.00
    • 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.

bloaty_fuzz_target summary

Ranking by median reached code coverage
Reached code coverage distribution
Mean code coverage growth over time
Mean code coverage growth over time
* The error bands show the 95% confidence interval around the mean code coverage.
  • Sample statistics and statistical significance (code coverage)
    Code coverage sample statistics
    count mean std min 25% median 75% max
    fuzzer time
    bandfuzz_syncmore 86400 10.0 6081.3 137.130959 5835.0 6059.25 6094.5 6156.00 6270.0
    aflplusplus 86400 10.0 5997.8 145.997565 5752.0 5869.25 6053.0 6109.75 6138.0
    bandfuzz_syncless 86400 10.0 5935.1 56.919924 5879.0 5897.00 5920.5 5951.75 6062.0

    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.
  • Unique code coverage plots
    Ranking by unique code branches covered
    Each bar shows the total number of code branches found by a given fuzzer. The colored area shows the number of unique code branches (i.e., branches that were not covered by any other fuzzers).
    Pairwise unique code coverage
    Each cell represents the number of code branches covered by the fuzzer of the column but not by the fuzzer of the row

curl_curl_fuzzer_http summary

Ranking by median reached code coverage
Reached code coverage distribution
Mean code coverage growth over time
Mean code coverage growth over time
* The error bands show the 95% confidence interval around the mean code coverage.
  • Sample statistics and statistical significance (code coverage)
    Code coverage sample statistics
    count mean std min 25% median 75% max
    fuzzer time
    bandfuzz_syncmore 86400 10.0 10795.9 80.039504 10674.0 10744.75 10799.5 10862.25 10917.0
    aflplusplus 86400 10.0 10764.9 35.501017 10698.0 10752.50 10770.5 10785.25 10809.0
    bandfuzz_syncless 86400 10.0 10783.6 94.130170 10698.0 10713.00 10746.0 10855.25 10940.0

    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.
  • Unique code coverage plots
    Ranking by unique code branches covered
    Each bar shows the total number of code branches found by a given fuzzer. The colored area shows the number of unique code branches (i.e., branches that were not covered by any other fuzzers).
    Pairwise unique code coverage
    Each cell represents the number of code branches covered by the fuzzer of the column but not by the fuzzer of the row

freetype2_ftfuzzer summary

Ranking by median reached code coverage
Reached code coverage distribution
Mean code coverage growth over time
Mean code coverage growth over time
* The error bands show the 95% confidence interval around the mean code coverage.
  • Sample statistics and statistical significance (code coverage)
    Code coverage sample statistics
    count mean std min 25% median 75% max
    fuzzer time
    bandfuzz_syncmore 86400 10.0 12300.9 450.261270 11611.0 11968.25 12509.5 12616.50 12862.0
    bandfuzz_syncless 86400 10.0 12291.9 648.798454 10934.0 11940.50 12430.0 12818.25 13000.0
    aflplusplus 86400 10.0 10851.7 560.688872 10141.0 10470.75 10677.5 11287.75 11786.0

    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.
  • Unique code coverage plots
    Ranking by unique code branches covered
    Each bar shows the total number of code branches found by a given fuzzer. The colored area shows the number of unique code branches (i.e., branches that were not covered by any other fuzzers).
    Pairwise unique code coverage
    Each cell represents the number of code branches covered by the fuzzer of the column but not by the fuzzer of the row

harfbuzz_hb-shape-fuzzer summary

Ranking by median reached code coverage
Reached code coverage distribution
Mean code coverage growth over time
Mean code coverage growth over time
* The error bands show the 95% confidence interval around the mean code coverage.
  • Sample statistics and statistical significance (code coverage)
    Code coverage sample statistics
    count mean std min 25% median 75% max
    fuzzer time
    aflplusplus 86400 10.0 10808.7 239.668312 10214.0 10838.00 10893.5 10943.00 10987.0
    bandfuzz_syncmore 86400 10.0 10892.6 21.782766 10849.0 10883.50 10893.0 10903.75 10924.0
    bandfuzz_syncless 86400 10.0 10895.0 51.523457 10836.0 10851.25 10876.5 10948.50 10965.0

    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.
  • Unique code coverage plots
    Ranking by unique code branches covered
    Each bar shows the total number of code branches found by a given fuzzer. The colored area shows the number of unique code branches (i.e., branches that were not covered by any other fuzzers).
    Pairwise unique code coverage
    Each cell represents the number of code branches covered by the fuzzer of the column but not by the fuzzer of the row

lcms_cms_transform_fuzzer summary

Ranking by median reached code coverage
Reached code coverage distribution
Mean code coverage growth over time
Mean code coverage growth over time
* The error bands show the 95% confidence interval around the mean code coverage.
  • Sample statistics and statistical significance (code coverage)
    Code coverage sample statistics
    count mean std min 25% median 75% max
    fuzzer time
    bandfuzz_syncmore 86400 10.0 2142.3 52.298396 2077.0 2098.50 2137.0 2189.50 2215.0
    bandfuzz_syncless 86400 10.0 2127.6 77.670529 1976.0 2098.25 2135.0 2181.75 2226.0
    aflplusplus 86400 10.0 1853.7 249.710966 1520.0 1586.75 1989.5 2062.00 2101.0

    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.
  • Unique code coverage plots
    Ranking by unique code branches covered
    Each bar shows the total number of code branches found by a given fuzzer. The colored area shows the number of unique code branches (i.e., branches that were not covered by any other fuzzers).
    Pairwise unique code coverage
    Each cell represents the number of code branches covered by the fuzzer of the column but not by the fuzzer of the row

libjpeg-turbo_libjpeg_turbo_fuzzer summary

Ranking by median reached code coverage
Reached code coverage distribution
Mean code coverage growth over time
Mean code coverage growth over time
* The error bands show the 95% confidence interval around the mean code coverage.
  • Sample statistics and statistical significance (code coverage)
    Code coverage sample statistics
    count mean std min 25% median 75% max
    fuzzer time
    aflplusplus 86400 10.0 2548.3 2.162817 2545.0 2547.0 2548.0 2550.00 2551.0
    bandfuzz_syncmore 86400 10.0 2546.9 1.911951 2545.0 2546.0 2546.0 2547.75 2551.0
    bandfuzz_syncless 86400 10.0 2546.6 2.366432 2544.0 2545.0 2545.5 2547.75 2551.0

    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.
  • Unique code coverage plots
    Ranking by unique code branches covered
    Each bar shows the total number of code branches found by a given fuzzer. The colored area shows the number of unique code branches (i.e., branches that were not covered by any other fuzzers).
    Pairwise unique code coverage
    Each cell represents the number of code branches covered by the fuzzer of the column but not by the fuzzer of the row

experiment data

You can download the raw data for this report here.

Check out the documentation on how to create customized reports using this data. Also see some example Colab notebooks for doing custom analysis on the data here.

The experiment was conducted using this FuzzBench commit: b690e5cb23d78fef3700e24b5df6dcd62de19103

To reproduce this experiment run the following commands in your FuzzBench repo:
# Check out the right commit.
git checkout b690e5cb23d78fef3700e24b5df6dcd62de19103
# Download the internal config file.
curl https://storage.googleapis.com//data/wenxuan/fb-data/exp/jan13-lab/input/config/experiment.yaml > /tmp/experiment-config.yaml
make install-dependencies
# Launch the experiment using paramters from the internal config file.
PYTHONPATH=. python experiment/reproduce_experiment.py -c /tmp/experiment-config.yaml -e <new_experiment_name>

Experiment Description:

(None,)