Published June 30, 2024 | Version v2
Dataset Open

Benchmarks for Comparing Binary-Level Symbolic Execution Speed

  • 1. University of Bremen
  • 1. Deutsches Forschungszentrum für Künstliche Intelligenz GmbH
  • 2. University of Bremen

Description

This dataset contains a reproducible computing environment as well as a set of synthetic benchmark applications for comparing symbolic execution speed at the binary level of several symbolic execution engines supporting the 32-bit RISC-V architecture. Specifically, the dataset compares and supports the following engines:

The benchmarks are designed to be easily extendable through the utilization of Guix (a functional package manager for long-term reproducible research). Currently, the supported engines perform IR-based symbolic execution using binary lifting (BinSec & angr) and direct symbolic execution of RISC-V machine code (SymEx-VP & BinSym). The dataset compares these approaches by fully exploring synthetic benchmark applications (e.g. sorting algorithms) with unconstrained symbolic input values and calculating an arithmetic mean over multiple executions, which is then represented as a table and a plot.

More information on using the dataset is available in the supplementary README file.

Files

symex-bench.zip

Files (1.6 GB)

Name Size Download all
md5:70a211c53a79513aee37f16170542c69
1.6 GB Preview Download

Additional details