KLEE is an open-source, symbolic execution-based, advanced fuzzing tool. The paper introducing it in OSDI 2008 is one of the most-cited works in the space of symbolic execution-based fuzzing. I wanted to see if I can reproduce their findings with the current version of KLEE and perform some additional experiments. This repository contains all code and results from these experiments.
WARNING: I kept fairly expensive records of each experiment, so this repository is >30GB.
This repository contains the following:
- Code to run the experiments
[*.]Dockerfile
: The setup for three different versions of coreutils.analyze.sh
: Script performing the actual analysis.run-suite.py
: Script orchestrating a run across all 89 coreutils.
- Results and result analysis
- out: The raw outputs from all 19 suite runs performed during this project
analyze.ipynb
: Scripts to analyze the raw outputs- plots: A series of plots on the results
- The report of the project
The report of this project can be found here.