Skip to content

A benchmarking suite for Joern

Notifications You must be signed in to change notification settings

joernio/joern-benchmarks

Repository files navigation

Joern Benchmarks

A repository for running Joern against known benchmarks.

Usage

$ sbt stage
$ ./joern-benchmarks --help
joern-benchmark v0.0.1
Usage: joern-benchmark [options] benchmark frontend

A benchmarking suite for Joern
  -h, --help
  --version                Prints the version
  benchmark                The benchmark to run. Available [SECURIBENCH_MICRO,ICHNAEA,THORAT]
  frontend                 The frontend to use. Available [JAVASRC,JAVA,JSSRC,PYSRC,SEMGREP,CODEQL]
  -d, --dataset-dir <value>
                           The dataset directory where benchmarks will be initialized and executed. Default is `./workspace`.
  -o, --output <value>     The output directory to write results to. Default is `./results`.
  -f, --format <value>     The output format to write results as. Default is MD. Available [JSON,CSV,MD]
  -k, --max-call-depth <value>
                           The max call depth `k` for the data-flow engine. Has no effect on non-Joern frontends. Default is 5.
  -i, --iterations <value>
                           The number of iterations for a given benchmark. Default is 1.

Example of testing for various values of k:

for k in {0..5}; do ./joern-benchmarks DEFECTS4J JAVA -f CSV -J-Xmx8G -i 1 -k $k; done 

Data-Flow Benchmarks

The benchmark naming convention of <BENCHMARK>_<FRONTEND>, e.g. OWASP_JAVA runs OWASP using the jimple2cpg frontend (JVM bytecode).

Benchmark Enabled Frontends
SECURIBENCH_MICRO JAVASRC JAVA SEMGREP CODEQL
ICHNAEA JSSRC SEMGREP
THORAT PYSRC SEMGREP CODEQL

Joern

Joern's open-source data-flow engine is enabled whenever a Joern frontend is selected, e.g. JAVA, PYSRC, etc.

Semgrep

If SEMGREP is selected, this requires an installation of Semgrep where semgrep scan will be used to initiate the process. Custom rules specific to benchmarks can be found under src/main/resources/semgrep.

Note: Only results with data-flow traces are considered as findings.

CodeQL

If CODEQL is selected, this requires an installation of CodeQL CLI where codeql will be used to create the database and run the scans. Custom rules specific to benchmarks can be found under src/main/resources/codeql.

Notes

Benchmarks successfully tested on the following versions of target software:

  • Joern v4.0.119
  • Semgrep v1.93.0
  • CodeQL v2.19.2

About

A benchmarking suite for Joern

Resources

Stars

Watchers

Forks

Packages

No packages published

Languages