Installation¶

Requirements¶

Requirement	Version
Docker	≥ 24
Docker Compose	≥ 2
Python	≥ 3.10 (non-Docker path only)
ezkl	22.0.1

Docker (recommended)¶

The fastest way to reproduce results is via Docker Compose, which bundles the utility (witness/prover), stats-logger, and regulator into pre-built images.

git clone https://github.com/disys-lab/zkSTAR.git
cd zkSTAR

Place your dataset CSV in data/ (see Configuration for the expected column layout), then follow the zkstarctl quickstart.

Python (development)¶

If you want to run or modify the training/evaluation code directly:

git clone https://github.com/disys-lab/zkSTAR.git
cd zkSTAR
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Key dependencies installed by requirements.txt:

Package	Role
`ezkl==22.0.1`	zkSNARK circuit compilation, witness generation, proving
`torch`	EKF-LSTM model and ONNX export
`onnx==1.18.0`	ONNX graph manipulation
`scipy`	Chi-squared CDF / UCL computation
`pandas`, `numpy`	Data loading and pre-processing
`scikit-learn`	MinMaxScaler normalisation
`PyYAML`	Config file parsing
`matplotlib`, `seaborn`	Result plotting

Building the documentation¶

pip install mkdocs-material mkdocstrings[python]
mkdocs serve        # live preview at http://127.0.0.1:8000
mkdocs build        # static site in site/