Setup

Create Virtual Environment

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
python <binaryninja_path>/scripts/install_api.py --install-on-pyenv

Initialize Submodules

git submodule update --init

Place Binaries

Place binary files in the dataset/bins/ directory and corresponding C source files in dataset/src/.

---

Usage

Pipeline

python pipeline.py

Options:

• -f, --fresh - Overwrite existing files
• -p, --processes - Number of parallel processes (default: CPU count)
• -d, --decomp-processes - Workers for decompilation
• -t, --decompile-timeout - Timeout per job in seconds (default: 600s)
• -m, --mem-limit - Memory limit per worker in GB (default: 10GB)
• -z, --decomp-mem-limit - Memory limit for decompilation in GB (default: 14GB)
• -g, --graph-edit-timeout - GED timeout in seconds (default: 870s)
• -s, --decomp-comp-threshold - Threshold for algorithm comparison

Individual Stages

You can run individual stages to continue an aborted run or with new limits:

Prepare Jobs:

python prepare_jobs.py

Options: -p (processes), -m (mem-limit)

Decompile:

python decomp_bins.py -f

Options: -f (fresh), -p (processes), -t (decompile-timeout), -m (mem-limit)

Process Functions:

python process_functions.py -f

Options: -f (fresh), -p (processes), -m (mem-limit)

Compute Results:

python comp_res.py -f

Options: -f (fresh), -p (processes), -m (mem-limit), -g (graph-edit-timeout), -s (decomp-comp-threshold)

Generate Plots:

python makePlots.py

Options: -r (ged-reconstruct-timeouts), -p (pathOverride)

Filter Interesting Functions:

python filterInterestingFunctions.py

Options: -p (pathOverride)

---

Dataset Structure

dataset/
├── bins/          # Place your binary files here
├── src/           # Place corresponding C source files here (.c)
├── jobs.json      # Function job queue (generated)
├── log.txt        # Pipeline log
├── decomp_log.txt # Decompilation log
├── conditional/   # Algorithm outputs
│   ├── decomp/           # Full decompiled functions
│   ├── decomp_func/      # Extracted decompiled functions
│   ├── src_func/         # Extracted source functions
│   ├── dict/             # SSA variable dictionaries (JSON)
│   ├── res/              # Comparison results
│   ├── plots/            # Visualizations
├── sreedhar/      # Algorithm outputs (same structure)
└── boissinot2008/ # Algorithm outputs (same structure)

---

Metrics

Graph Edit Distance (GED)

• total_ged - Total graph edit distance
• matched_ged - GED for matched components
• unmatched_ged - GED for unmatched components

Structural Metrics

• total_nodes1 - Node count src graph
• total_nodes2 - Node count decomp graph
• anz_zhk1 - Number of ZHKs in src grpah
• anz_zhk2 - Number of ZHKs in decomp graph
• matched_zhk - Exact ZHK matches
• halfmatched_zhk - ZHK matches

Performance

• time_needed - Processing time
• runs - Successful runs