lattice-estimator-cli: CLI for Lattice Estimator

This CLI is for calling lattice estimator. It estimates the security parameter from ROP for the given parameters.

Usage

python -m lattice_cli <ring_dim> <q> \
  --s-dist '{"name": "DiscreteGaussianAlpha", "alpha": 0.001, "q": 12289}' \
  --e-dist '{"name": "CenteredBinomial", "eta": 3}' \
  [--m <int>] [--exact]

• <ring_dim>: ring dimension (n)
• <q>: modulus q
• --s-dist: JSON spec for the secret noise distribution
• --e-dist: JSON spec for the error noise distribution
• --m: optional number of samples; if omitted, infinity is used
• --exact: use exact estimation; by default, a rough estimate is used (The CLI automatically uses the bundled estimator/ submodule in this repo.)

Distribution specs

Provide a JSON object with at least a name field. Supported types and fields:

• DiscreteGaussian: { "stddev": float, "mean"?: float, "n"?: int }
• DiscreteGaussianAlpha: { "alpha": float, "q"?: int, "mean"?: float, "n"?: int }
• CenteredBinomial: { "eta": int, "n"?: int }
• Uniform: { "a": int, "b": int, "n"?: int }
• UniformMod: { "q": int, "n"?: int }
• SparseTernary: { "p": int, "m": int, "n"?: int }
• SparseBinary: { "hw": int, "n"?: int }
• Binary: {} (alias of Uniform(0,1))`
• Ternary: {} (alias of Uniform(-1,1))`

Notes:

• For DiscreteGaussianAlpha and UniformMod, if q is omitted in the spec, the top-level <q> argument is used.

Examples

# Rough estimate (default)
python -m lattice_cli 1024 12289 \
  --s-dist '{"name": "DiscreteGaussianAlpha", "alpha": 0.001}' \
  --e-dist '{"name": "CenteredBinomial", "eta": 3}'

# Exact estimate with explicit m
python -m lattice_cli 1024 12289 \
  --s-dist '{"name": "Binary"}' \
  --e-dist '{"name": "DiscreteGaussian", "stddev": 3.2}' \
  --m 100000 --exact

Run From Anywhere

Use the provided wrapper so the local estimator/ submodule is found and Sage's Python is used:

# From the repository root
./scripts/lattice-estimator-cli 1024 12289 --s-dist '{"name":"CenteredBinomial","eta":3}' --e-dist '{"name":"Ternary"}'

# Or put it on PATH
export PATH="$PWD/scripts:$PATH"
lattice-estimator-cli 1024 12289 --s-dist '{"name":"CenteredBinomial","eta":3}' --e-dist '{"name":"Ternary"}'

The wrapper sets PYTHONPATH to the repository root and executes sage -python -m lattice_cli ....

Call From Rust (example)

use std::process::Command;

fn main() {
    let output = Command::new("/absolute/path/to/repo/scripts/lattice-estimator-cli")
        .args([
            "1024",
            "12289",
            "--s-dist",
            "{\"name\":\"CenteredBinomial\",\"eta\":3}",
            "--e-dist",
            "{\"name\":\"Ternary\"}",
        ])
        .output()
        .expect("failed to execute process");

    if output.status.success() {
        let stdout = String::from_utf8_lossy(&output.stdout);
        println!("secpar = {}", stdout.trim());
    } else {
        eprintln!("stderr: {}", String::from_utf8_lossy(&output.stderr));
    }
}

Repository layout

The CLI lives under the lattice_cli/ package and can be invoked with python -m lattice_cli.

• lattice_cli/cli.py: Main CLI implementation (argparse‑based)
• lattice_cli/__main__.py: Bootstraps python -m lattice_cli
• lattice_cli/__init__.py: Package metadata (version, etc.)
• lattice_cli/core.py: The estimate_rop_secpar implementation

Notes on dependencies

The CLI automatically adds the local estimator/ submodule (at repo root) to sys.path. Some estimator modules depend on Sage (sage); ensure Sage and other dependencies are installed in your environment.