provael 0.1.0__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- provael-0.1.0/.github/ISSUE_TEMPLATE/config.yml +5 -0
- provael-0.1.0/.github/ISSUE_TEMPLATE/leaderboard-submission.yml +49 -0
- provael-0.1.0/.github/PULL_REQUEST_TEMPLATE.md +20 -0
- provael-0.1.0/.github/workflows/ci.yml +34 -0
- provael-0.1.0/.github/workflows/leaderboard-submission.yml +28 -0
- provael-0.1.0/.github/workflows/release.yml +96 -0
- provael-0.1.0/.gitignore +53 -0
- provael-0.1.0/CHANGELOG.md +51 -0
- provael-0.1.0/CONTRIBUTING-leaderboard.md +48 -0
- provael-0.1.0/CONTRIBUTING.md +63 -0
- provael-0.1.0/LICENSE +201 -0
- provael-0.1.0/PKG-INFO +232 -0
- provael-0.1.0/PRIOR_ART.md +80 -0
- provael-0.1.0/README.md +203 -0
- provael-0.1.0/SAFETY.md +67 -0
- provael-0.1.0/SECURITY.md +29 -0
- provael-0.1.0/docs/assets/demo.svg +110 -0
- provael-0.1.0/leaderboard/README.md +71 -0
- provael-0.1.0/leaderboard/app.py +129 -0
- provael-0.1.0/leaderboard/requirements.txt +5 -0
- provael-0.1.0/leaderboard/results/leaderboard.json +80 -0
- provael-0.1.0/pyproject.toml +104 -0
- provael-0.1.0/results/smolvla_libero_object/README.md +25 -0
- provael-0.1.0/results/smolvla_libero_object/report.json +1197 -0
- provael-0.1.0/scripts/calibrate_zones.py +92 -0
- provael-0.1.0/scripts/record_demo.sh +54 -0
- provael-0.1.0/scripts/run_real.sh +76 -0
- provael-0.1.0/scripts/validate_submission.py +53 -0
- provael-0.1.0/src/provael/__init__.py +13 -0
- provael-0.1.0/src/provael/attacks/__init__.py +1 -0
- provael-0.1.0/src/provael/attacks/_image.py +77 -0
- provael-0.1.0/src/provael/attacks/_payload.py +51 -0
- provael-0.1.0/src/provael/attacks/base.py +50 -0
- provael-0.1.0/src/provael/attacks/baseline.py +28 -0
- provael-0.1.0/src/provael/attacks/injection.py +102 -0
- provael-0.1.0/src/provael/attacks/instruction.py +103 -0
- provael-0.1.0/src/provael/attacks/registry.py +120 -0
- provael-0.1.0/src/provael/attacks/visual.py +89 -0
- provael-0.1.0/src/provael/cli.py +250 -0
- provael-0.1.0/src/provael/config.py +50 -0
- provael-0.1.0/src/provael/leaderboard.py +215 -0
- provael-0.1.0/src/provael/policies/__init__.py +1 -0
- provael-0.1.0/src/provael/policies/base.py +57 -0
- provael-0.1.0/src/provael/policies/lerobot_adapter.py +260 -0
- provael-0.1.0/src/provael/policies/registry.py +94 -0
- provael-0.1.0/src/provael/policies/stub.py +134 -0
- provael-0.1.0/src/provael/py.typed +0 -0
- provael-0.1.0/src/provael/report.py +148 -0
- provael-0.1.0/src/provael/runner.py +151 -0
- provael-0.1.0/src/provael/scoring/__init__.py +1 -0
- provael-0.1.0/src/provael/scoring/asr.py +87 -0
- provael-0.1.0/src/provael/suites/__init__.py +70 -0
- provael-0.1.0/src/provael/suites/base.py +64 -0
- provael-0.1.0/src/provael/suites/keepout_zones.py +133 -0
- provael-0.1.0/src/provael/suites/libero.py +402 -0
- provael-0.1.0/src/provael/suites/stub.py +127 -0
- provael-0.1.0/src/provael/types.py +131 -0
- provael-0.1.0/tests/conftest.py +20 -0
- provael-0.1.0/tests/test_attacks.py +94 -0
- provael-0.1.0/tests/test_cli.py +125 -0
- provael-0.1.0/tests/test_determinism.py +39 -0
- provael-0.1.0/tests/test_keepout_zones.py +76 -0
- provael-0.1.0/tests/test_leaderboard.py +102 -0
- provael-0.1.0/tests/test_lerobot_adapter.py +155 -0
- provael-0.1.0/tests/test_libero_adapter.py +209 -0
- provael-0.1.0/tests/test_real_policy_injection.py +169 -0
- provael-0.1.0/tests/test_registries.py +63 -0
- provael-0.1.0/tests/test_runner_smoke.py +63 -0
- provael-0.1.0/tests/test_scoring.py +83 -0
- provael-0.1.0/tests/test_validate_submission.py +38 -0
- provael-0.1.0/tests/test_visual_injection.py +181 -0
- provael-0.1.0/uv.lock +3004 -0
|
@@ -0,0 +1,5 @@
|
|
|
1
|
+
blank_issues_enabled: true
|
|
2
|
+
contact_links:
|
|
3
|
+
- name: Security vulnerability (report privately)
|
|
4
|
+
url: https://github.com/provael/provael/security/advisories/new
|
|
5
|
+
about: Please report security issues privately via a GitHub security advisory — see SECURITY.md, not a public issue.
|
|
@@ -0,0 +1,49 @@
|
|
|
1
|
+
name: Leaderboard submission
|
|
2
|
+
description: Submit a VLA policy's robustness result to the Provael leaderboard
|
|
3
|
+
title: "[leaderboard] <policy> on <suite>"
|
|
4
|
+
labels: ["leaderboard"]
|
|
5
|
+
body:
|
|
6
|
+
- type: markdown
|
|
7
|
+
attributes:
|
|
8
|
+
value: |
|
|
9
|
+
Thanks for contributing a result! The fastest path is a **pull request** adding
|
|
10
|
+
`results/<name>/report.json` — see
|
|
11
|
+
[CONTRIBUTING-leaderboard.md](../blob/main/CONTRIBUTING-leaderboard.md). Use this
|
|
12
|
+
issue if you'd like help, can't open a PR, or want to propose a checkpoint to add.
|
|
13
|
+
- type: input
|
|
14
|
+
id: policy
|
|
15
|
+
attributes:
|
|
16
|
+
label: Policy / checkpoint
|
|
17
|
+
placeholder: "e.g. HuggingFaceVLA/smolvla_libero"
|
|
18
|
+
validations:
|
|
19
|
+
required: true
|
|
20
|
+
- type: input
|
|
21
|
+
id: suite
|
|
22
|
+
attributes:
|
|
23
|
+
label: Suite + task(s)
|
|
24
|
+
placeholder: "e.g. libero_object/0,1,2"
|
|
25
|
+
validations:
|
|
26
|
+
required: true
|
|
27
|
+
- type: input
|
|
28
|
+
id: setup
|
|
29
|
+
attributes:
|
|
30
|
+
label: Seeds / horizon / hardware
|
|
31
|
+
placeholder: "e.g. 10 seeds, horizon 280, RTX 4090, MUJOCO_GL=osmesa"
|
|
32
|
+
validations:
|
|
33
|
+
required: true
|
|
34
|
+
- type: textarea
|
|
35
|
+
id: report
|
|
36
|
+
attributes:
|
|
37
|
+
label: report.json (paste or link)
|
|
38
|
+
description: Paste the `report.json` produced by `provael attack`, or link the run.
|
|
39
|
+
validations:
|
|
40
|
+
required: true
|
|
41
|
+
- type: checkboxes
|
|
42
|
+
id: checks
|
|
43
|
+
attributes:
|
|
44
|
+
label: Checklist
|
|
45
|
+
options:
|
|
46
|
+
- label: "Produced via `provael attack` and includes a `none` baseline"
|
|
47
|
+
required: true
|
|
48
|
+
- label: "I understand results are seeded/reproducible and may be re-run"
|
|
49
|
+
required: true
|
|
@@ -0,0 +1,20 @@
|
|
|
1
|
+
## Summary
|
|
2
|
+
|
|
3
|
+
<!-- What does this change and why? -->
|
|
4
|
+
|
|
5
|
+
## Test plan
|
|
6
|
+
|
|
7
|
+
<!-- How did you verify it? For code changes: -->
|
|
8
|
+
<!-- `uv run ruff check . && uv run mypy src && uv run pytest -q` -->
|
|
9
|
+
|
|
10
|
+
---
|
|
11
|
+
|
|
12
|
+
### Leaderboard submission? (delete this section if N/A)
|
|
13
|
+
|
|
14
|
+
- [ ] Adds `results/<name>/report.json` produced by `provael attack` (with a `none` baseline)
|
|
15
|
+
- [ ] `python scripts/validate_submission.py` passes locally
|
|
16
|
+
- [ ] Reproducible: checkpoint, suite/task(s), seeds, horizon, and hardware noted below
|
|
17
|
+
|
|
18
|
+
<!-- policy/checkpoint:
|
|
19
|
+
suite/tasks:
|
|
20
|
+
seeds / horizon / hardware: -->
|
|
@@ -0,0 +1,34 @@
|
|
|
1
|
+
name: CI
|
|
2
|
+
|
|
3
|
+
on:
|
|
4
|
+
push:
|
|
5
|
+
branches: [main]
|
|
6
|
+
pull_request:
|
|
7
|
+
|
|
8
|
+
# CPU-only by design: this workflow NEVER installs the optional `[lerobot]` extra,
|
|
9
|
+
# so it pulls no torch / transformers / GPU stack. The LeRobot integration tests are
|
|
10
|
+
# gated (PROVAEL_INTEGRATION) and skip here automatically.
|
|
11
|
+
jobs:
|
|
12
|
+
check:
|
|
13
|
+
name: ruff + mypy + pytest (CPU, no lerobot)
|
|
14
|
+
runs-on: ubuntu-latest
|
|
15
|
+
steps:
|
|
16
|
+
- uses: actions/checkout@v4
|
|
17
|
+
|
|
18
|
+
- name: Install uv (pinned Python 3.12)
|
|
19
|
+
uses: astral-sh/setup-uv@v5
|
|
20
|
+
with:
|
|
21
|
+
python-version: "3.12"
|
|
22
|
+
enable-cache: true
|
|
23
|
+
|
|
24
|
+
- name: Sync core + dev (no extras)
|
|
25
|
+
run: uv sync --locked
|
|
26
|
+
|
|
27
|
+
- name: Lint (ruff)
|
|
28
|
+
run: uv run ruff check .
|
|
29
|
+
|
|
30
|
+
- name: Type-check (mypy, strict)
|
|
31
|
+
run: uv run mypy src
|
|
32
|
+
|
|
33
|
+
- name: Test (pytest, CPU only)
|
|
34
|
+
run: uv run pytest -q
|
|
@@ -0,0 +1,28 @@
|
|
|
1
|
+
name: Leaderboard submission
|
|
2
|
+
|
|
3
|
+
# Validates external leaderboard submissions on PRs that add/change results/**.
|
|
4
|
+
# CPU-only — never installs lerobot. Confirms each report.json is a valid, internally
|
|
5
|
+
# consistent RunReport and that the leaderboard still builds from results/*.
|
|
6
|
+
|
|
7
|
+
on:
|
|
8
|
+
pull_request:
|
|
9
|
+
paths: ["results/**"]
|
|
10
|
+
|
|
11
|
+
permissions: {} # read-only; least privilege
|
|
12
|
+
|
|
13
|
+
jobs:
|
|
14
|
+
validate:
|
|
15
|
+
name: Validate submission + rebuild leaderboard
|
|
16
|
+
runs-on: ubuntu-latest
|
|
17
|
+
steps:
|
|
18
|
+
- uses: actions/checkout@v4
|
|
19
|
+
- name: Install uv (Python 3.12)
|
|
20
|
+
uses: astral-sh/setup-uv@v5
|
|
21
|
+
with:
|
|
22
|
+
python-version: "3.12"
|
|
23
|
+
- name: Install (locked, CPU core only)
|
|
24
|
+
run: uv sync --locked
|
|
25
|
+
- name: Validate submitted report(s)
|
|
26
|
+
run: uv run python scripts/validate_submission.py 'results/*'
|
|
27
|
+
- name: Leaderboard builds cleanly
|
|
28
|
+
run: uv run provael leaderboard build --runs 'results/*' --out "${RUNNER_TEMP}/lb"
|
|
@@ -0,0 +1,96 @@
|
|
|
1
|
+
name: Release
|
|
2
|
+
|
|
3
|
+
# Tag-triggered release: build sdist + wheel, publish to PyPI via OIDC TRUSTED
|
|
4
|
+
# PUBLISHING (no API token stored anywhere), then cut a GitHub Release from the
|
|
5
|
+
# CHANGELOG.
|
|
6
|
+
#
|
|
7
|
+
# ────────────────────────────────────────────────────────────────────────────
|
|
8
|
+
# ONE-TIME MAINTAINER SETUP (PyPI trusted publisher — no secrets in the repo):
|
|
9
|
+
# 1. Create the project owner account on https://pypi.org.
|
|
10
|
+
# 2. PyPI → your account → "Publishing" → "Add a new pending publisher":
|
|
11
|
+
# PyPI project name : provael
|
|
12
|
+
# Owner : provael
|
|
13
|
+
# Repository name : provael
|
|
14
|
+
# Workflow name : release.yml
|
|
15
|
+
# Environment name : pypi
|
|
16
|
+
# 3. In GitHub → repo Settings → Environments → create an environment named "pypi"
|
|
17
|
+
# (optionally add required reviewers / a tag protection rule).
|
|
18
|
+
# 4. Publish by pushing a tag that matches the version, e.g.:
|
|
19
|
+
# git tag v0.1.0 && git push origin v0.1.0
|
|
20
|
+
# (For a first dry run, point step `publish` at TestPyPI by setting
|
|
21
|
+
# `repository-url: https://test.pypi.org/legacy/` and adding the matching pending
|
|
22
|
+
# publisher on test.pypi.org.)
|
|
23
|
+
# ────────────────────────────────────────────────────────────────────────────
|
|
24
|
+
|
|
25
|
+
on:
|
|
26
|
+
push:
|
|
27
|
+
tags: ["v*"]
|
|
28
|
+
|
|
29
|
+
permissions: {} # least privilege by default; jobs opt into exactly what they need
|
|
30
|
+
|
|
31
|
+
jobs:
|
|
32
|
+
build:
|
|
33
|
+
name: Build sdist + wheel
|
|
34
|
+
runs-on: ubuntu-latest
|
|
35
|
+
steps:
|
|
36
|
+
- uses: actions/checkout@v4
|
|
37
|
+
- name: Install uv (Python 3.12)
|
|
38
|
+
uses: astral-sh/setup-uv@v5
|
|
39
|
+
with:
|
|
40
|
+
python-version: "3.12"
|
|
41
|
+
- name: Build distributions
|
|
42
|
+
run: uv build
|
|
43
|
+
- name: Check metadata
|
|
44
|
+
run: uvx twine check dist/*
|
|
45
|
+
- uses: actions/upload-artifact@v4
|
|
46
|
+
with:
|
|
47
|
+
name: dist
|
|
48
|
+
path: dist/
|
|
49
|
+
|
|
50
|
+
publish-pypi:
|
|
51
|
+
name: Publish to PyPI (OIDC trusted publishing)
|
|
52
|
+
needs: build
|
|
53
|
+
runs-on: ubuntu-latest
|
|
54
|
+
environment: pypi
|
|
55
|
+
permissions:
|
|
56
|
+
id-token: write # REQUIRED for OIDC trusted publishing — no token needed
|
|
57
|
+
steps:
|
|
58
|
+
- uses: actions/download-artifact@v4
|
|
59
|
+
with:
|
|
60
|
+
name: dist
|
|
61
|
+
path: dist/
|
|
62
|
+
- name: Publish
|
|
63
|
+
uses: pypa/gh-action-pypi-publish@release/v1
|
|
64
|
+
with:
|
|
65
|
+
skip-existing: true # don't hard-fail if a re-run hits an already-uploaded version
|
|
66
|
+
# No `password:` — trusted publishing uses the OIDC id-token above.
|
|
67
|
+
|
|
68
|
+
github-release:
|
|
69
|
+
name: Create GitHub Release
|
|
70
|
+
needs: publish-pypi
|
|
71
|
+
runs-on: ubuntu-latest
|
|
72
|
+
permissions:
|
|
73
|
+
contents: write # to create the release
|
|
74
|
+
steps:
|
|
75
|
+
- uses: actions/checkout@v4
|
|
76
|
+
- uses: actions/download-artifact@v4
|
|
77
|
+
with:
|
|
78
|
+
name: dist
|
|
79
|
+
path: dist/
|
|
80
|
+
- name: Extract changelog notes for this tag
|
|
81
|
+
id: notes
|
|
82
|
+
run: |
|
|
83
|
+
version="${GITHUB_REF_NAME#v}"
|
|
84
|
+
awk -v ver="$version" '
|
|
85
|
+
$0 ~ "^## \\[" ver "\\]" {flag=1; next}
|
|
86
|
+
flag && /^## \[/ {exit}
|
|
87
|
+
flag {print}
|
|
88
|
+
' CHANGELOG.md > release-notes.md
|
|
89
|
+
if [ ! -s release-notes.md ]; then echo "Release ${GITHUB_REF_NAME}" > release-notes.md; fi
|
|
90
|
+
- name: Create release
|
|
91
|
+
env:
|
|
92
|
+
GH_TOKEN: ${{ github.token }}
|
|
93
|
+
run: |
|
|
94
|
+
gh release create "${GITHUB_REF_NAME}" dist/* \
|
|
95
|
+
--title "${GITHUB_REF_NAME}" \
|
|
96
|
+
--notes-file release-notes.md
|
provael-0.1.0/.gitignore
ADDED
|
@@ -0,0 +1,53 @@
|
|
|
1
|
+
# Python
|
|
2
|
+
__pycache__/
|
|
3
|
+
*.py[cod]
|
|
4
|
+
*$py.class
|
|
5
|
+
*.so
|
|
6
|
+
.Python
|
|
7
|
+
build/
|
|
8
|
+
dist/
|
|
9
|
+
*.egg-info/
|
|
10
|
+
.eggs/
|
|
11
|
+
develop-eggs/
|
|
12
|
+
.installed.cfg
|
|
13
|
+
*.egg
|
|
14
|
+
|
|
15
|
+
# Virtual environments
|
|
16
|
+
.venv/
|
|
17
|
+
venv/
|
|
18
|
+
env/
|
|
19
|
+
ENV/
|
|
20
|
+
.uv/
|
|
21
|
+
|
|
22
|
+
# uv
|
|
23
|
+
uv.lock.bak
|
|
24
|
+
|
|
25
|
+
# Testing / coverage
|
|
26
|
+
.pytest_cache/
|
|
27
|
+
.coverage
|
|
28
|
+
.coverage.*
|
|
29
|
+
coverage.xml
|
|
30
|
+
htmlcov/
|
|
31
|
+
.mypy_cache/
|
|
32
|
+
.ruff_cache/
|
|
33
|
+
.hypothesis/
|
|
34
|
+
|
|
35
|
+
# Run artifacts (generated reports)
|
|
36
|
+
runs/
|
|
37
|
+
|
|
38
|
+
# Optional demo recording (regenerated by scripts/record_demo.sh when asciinema+agg
|
|
39
|
+
# are installed); the committed deterministic asset is docs/assets/demo.svg.
|
|
40
|
+
docs/assets/*.gif
|
|
41
|
+
|
|
42
|
+
# OS / editor
|
|
43
|
+
.DS_Store
|
|
44
|
+
.idea/
|
|
45
|
+
.vscode/
|
|
46
|
+
*.swp
|
|
47
|
+
*~
|
|
48
|
+
|
|
49
|
+
# Jupyter
|
|
50
|
+
.ipynb_checkpoints/
|
|
51
|
+
|
|
52
|
+
# private strategy/market docs — never publish
|
|
53
|
+
/vla-redteam-*.md
|
|
@@ -0,0 +1,51 @@
|
|
|
1
|
+
# Changelog
|
|
2
|
+
|
|
3
|
+
All notable changes to this project are documented here. The format is based on
|
|
4
|
+
[Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and this project adheres to
|
|
5
|
+
[Semantic Versioning](https://semver.org/spec/v2.0.0.html).
|
|
6
|
+
|
|
7
|
+
## [0.1.0] — 2026-06-27
|
|
8
|
+
|
|
9
|
+
Initial **Provael** release. Provael is the open-source red-team & assurance layer for
|
|
10
|
+
physical AI — a CPU-first, model-agnostic harness that perturbs the instructions and
|
|
11
|
+
observations a Vision-Language-Action (VLA) robot policy receives in simulation and reports
|
|
12
|
+
an Attack Success Rate (ASR).
|
|
13
|
+
|
|
14
|
+
> **Renamed from `vla-redteam` / `robopwn`.** The same engine and full git history, under a
|
|
15
|
+
> new name, CLI (`provael`), and home (github.com/provael/provael). It was previously
|
|
16
|
+
> published on PyPI as `vla-redteam` (0.1.0–0.2.2); the env-var gate is now
|
|
17
|
+
> `PROVAEL_INTEGRATION`. Behavior is unchanged — the deterministic stub run still reports
|
|
18
|
+
> ASR 67.1% (47/70), byte-identical.
|
|
19
|
+
|
|
20
|
+
### Included
|
|
21
|
+
- **CPU-first deterministic core** — `StubPolicy` + `StubSuite` give exact, byte-reproducible
|
|
22
|
+
ASRs with no GPU or model download; strict typing, `py.typed`, 100 tests.
|
|
23
|
+
- **Three templated attack families** (`instruction`, `visual`, `injection`) + a `none`
|
|
24
|
+
baseline; ASR with per-attack / per-task breakdowns and seeded mean ± std for real policies.
|
|
25
|
+
- **Real SmolVLA × LIBERO path** behind the `[lerobot]` extra + `PROVAEL_INTEGRATION=1`,
|
|
26
|
+
replicating LeRobot's verified evaluator rollout; first real result on `libero_object/0`
|
|
27
|
+
(instruction attacks 60–100% vs a 0% benign baseline; visual/injection 0% — honest null).
|
|
28
|
+
- **Per-task calibrated keep-out zones** scaffold (`suites/keepout_zones.py` +
|
|
29
|
+
`scripts/calibrate_zones.py`) toward a calibrated hazard rate.
|
|
30
|
+
- **Leaderboard** — deterministic `(policy × suite × family) → ASR` table, a Gradio Space, and
|
|
31
|
+
a public submission flow (validator + CI on `results/**`).
|
|
32
|
+
- **CLI `provael`** — `attack`, `report`, `leaderboard build`, `list-policies`,
|
|
33
|
+
`list-attacks`, `version`; actionable errors (exit code 2), never a traceback.
|
|
34
|
+
- Apache-2.0; OIDC trusted-publishing release pipeline; strict `ruff` + `mypy` gate in CI.
|
|
35
|
+
|
|
36
|
+
### Honest scope
|
|
37
|
+
Attacks are templated (not gradient/optimization-based); only the instruction family
|
|
38
|
+
transfers to real SmolVLA so far; one policy + one suite shipped; the headline result uses a
|
|
39
|
+
single task with a default, uncalibrated keep-out predicate. See the README's
|
|
40
|
+
"Scope and honest limitations."
|
|
41
|
+
|
|
42
|
+
### Roadmap
|
|
43
|
+
- **v0.2.0** — SARIF output (`provael report --format sarif`) for GitHub code scanning; a
|
|
44
|
+
reusable `provael/provael-action` CI gate; an Embodied-AI Top-10 risk mapping; per-task
|
|
45
|
+
keep-out-zone calibration.
|
|
46
|
+
- **later** — optimized (gradient/search) attacks; a second policy/suite backend.
|
|
47
|
+
|
|
48
|
+
> Detailed pre-rebrand history (the `vla-redteam` 0.2.x line) is preserved in the git log and
|
|
49
|
+
> the prior PyPI releases.
|
|
50
|
+
|
|
51
|
+
[0.1.0]: https://github.com/provael/provael/releases/tag/v0.1.0
|
|
@@ -0,0 +1,48 @@
|
|
|
1
|
+
# Submitting to the Provael leaderboard
|
|
2
|
+
|
|
3
|
+
The [leaderboard](https://huggingface.co/spaces/Sattyam/provael-leaderboard) aggregates
|
|
4
|
+
real `(policy × suite × family) → ASR` results. Anyone can add a result via a pull request —
|
|
5
|
+
CI validates it automatically.
|
|
6
|
+
|
|
7
|
+
## How to submit (pull request)
|
|
8
|
+
|
|
9
|
+
1. **Run an attack** on your policy, including the `none` baseline (so we can read *lift*):
|
|
10
|
+
|
|
11
|
+
```bash
|
|
12
|
+
pip install 'provael[lerobot]' 'lerobot[libero]==0.5.1'
|
|
13
|
+
export MUJOCO_GL=osmesa PYOPENGL_PLATFORM=osmesa # headless rendering
|
|
14
|
+
provael attack --policy smolvla --suite libero --model <your-checkpoint> \
|
|
15
|
+
--attacks none,instruction,visual,injection --seeds 10 --horizon 280 --seed 0 \
|
|
16
|
+
--out results/<your-name> # writes results/<your-name>/report.json
|
|
17
|
+
```
|
|
18
|
+
|
|
19
|
+
2. **Validate locally** (same check CI runs):
|
|
20
|
+
|
|
21
|
+
```bash
|
|
22
|
+
python scripts/validate_submission.py 'results/*'
|
|
23
|
+
```
|
|
24
|
+
|
|
25
|
+
3. **Open a PR** adding `results/<your-name>/report.json`. The **Leaderboard submission**
|
|
26
|
+
workflow validates the report and confirms the leaderboard still builds. In the PR
|
|
27
|
+
description, note the checkpoint, suite/task(s), seeds, horizon, and hardware.
|
|
28
|
+
|
|
29
|
+
4. On merge, a maintainer rebuilds the published board
|
|
30
|
+
(`provael leaderboard build --runs 'results/*' --out leaderboard/results`) and the Space
|
|
31
|
+
updates. (`is_demo` clears automatically once any non-stub result is present.)
|
|
32
|
+
|
|
33
|
+
## What the validator checks
|
|
34
|
+
|
|
35
|
+
- The file parses as a `RunReport`.
|
|
36
|
+
- ASR ∈ [0, 1]; `successes` ∈ [0, `attempts`].
|
|
37
|
+
- `attempts` equals the number of **applicable** episodes (not-applicable attacks are
|
|
38
|
+
excluded from the denominator, never faked).
|
|
39
|
+
- `successes` matches the applicable successes in the per-episode results.
|
|
40
|
+
- Every result has an `attack` and `family`.
|
|
41
|
+
|
|
42
|
+
## Norms (please)
|
|
43
|
+
|
|
44
|
+
- **Reproducible & seeded.** Report the exact checkpoint, suite/tasks, seeds, and horizon.
|
|
45
|
+
- **Honest scope.** If you used a custom keep-out predicate or calibrated zones, say so.
|
|
46
|
+
- **No fabricated numbers.** Results may be independently re-run.
|
|
47
|
+
|
|
48
|
+
Questions or can't open a PR? Use the **Leaderboard submission** issue template.
|
|
@@ -0,0 +1,63 @@
|
|
|
1
|
+
# Contributing to provael
|
|
2
|
+
|
|
3
|
+
Thanks for your interest in improving **provael** (Provael). This is a defensive,
|
|
4
|
+
sim-only security-research tool — please read [SAFETY.md](SAFETY.md) first.
|
|
5
|
+
|
|
6
|
+
## Ground rules
|
|
7
|
+
|
|
8
|
+
- **CPU-first.** The entire core runs and is tested on a plain CPU with **no GPU and no
|
|
9
|
+
model/dataset download**. Real policies (SmolVLA via LeRobot) and the LIBERO simulator
|
|
10
|
+
live behind the optional `[lerobot]` extra and a `PROVAEL_INTEGRATION=1` gate. **CI never
|
|
11
|
+
installs `lerobot`** — keep it that way.
|
|
12
|
+
- **Verify, don't fabricate.** Numbers in the README/CHANGELOG must come from a real run.
|
|
13
|
+
When a third-party API is involved, read/introspect the installed source rather than
|
|
14
|
+
guessing; if something can't be verified (no GPU/sim), say so and ship a clearly-skipped
|
|
15
|
+
gated test instead of a guess.
|
|
16
|
+
- **Determinism.** The stub policy/suite are byte-deterministic; real policies are seeded
|
|
17
|
+
but model-stochastic (reported as mean ± per-seed std). Don't put wall-clock into
|
|
18
|
+
`report.json`.
|
|
19
|
+
|
|
20
|
+
## Dev setup
|
|
21
|
+
|
|
22
|
+
```bash
|
|
23
|
+
# uv (recommended): https://docs.astral.sh/uv/
|
|
24
|
+
uv sync # core + dev tools (no GPU stack)
|
|
25
|
+
uv run provael --help
|
|
26
|
+
```
|
|
27
|
+
|
|
28
|
+
## The gate (must be green before you push)
|
|
29
|
+
|
|
30
|
+
```bash
|
|
31
|
+
uv run ruff check . # lint + import order
|
|
32
|
+
uv run mypy src # strict type-check
|
|
33
|
+
uv run pytest -q # tests (GPU/LIBERO tests auto-skip without the extra)
|
|
34
|
+
```
|
|
35
|
+
|
|
36
|
+
The optional, GPU-gated integration tests:
|
|
37
|
+
|
|
38
|
+
```bash
|
|
39
|
+
pip install 'provael[lerobot]' 'lerobot[libero]==0.5.1'
|
|
40
|
+
PROVAEL_INTEGRATION=1 pytest tests/test_lerobot_adapter.py tests/test_libero_adapter.py -q
|
|
41
|
+
```
|
|
42
|
+
|
|
43
|
+
## Conventions
|
|
44
|
+
|
|
45
|
+
- **Python 3.12+**, type hints on all signatures, pydantic v2 for data models.
|
|
46
|
+
- **Commits**: imperative, conventional-style prefixes (`feat:`, `fix:`, `docs:`,
|
|
47
|
+
`test:`, `chore:`), concise subject (< 72 chars), body for context.
|
|
48
|
+
- **Architecture**: ports-and-adapters — new policies subclass `PolicyAdapter`, new suites
|
|
49
|
+
`SuiteAdapter`, new attacks `Attack`; register them in the relevant registry and add a
|
|
50
|
+
CPU test (use the stub) plus a gated test if a GPU/sim is required.
|
|
51
|
+
- Keep optional dependencies imported **only inside methods**, never at module scope.
|
|
52
|
+
|
|
53
|
+
## Pull requests
|
|
54
|
+
|
|
55
|
+
1. Branch (`feature/…`, `fix/…`), make the change, keep the gate green.
|
|
56
|
+
2. Add/extend tests; for new attacks, include the deterministic stub ASR.
|
|
57
|
+
3. Update `CHANGELOG.md` (`[Unreleased]`) and any affected docs.
|
|
58
|
+
4. Open the PR with a **Summary** and a **Test Plan**.
|
|
59
|
+
|
|
60
|
+
## Reporting bugs / requesting features
|
|
61
|
+
|
|
62
|
+
Open a GitHub issue (templates provided). For security vulnerabilities, see
|
|
63
|
+
[SECURITY.md](SECURITY.md) — please do **not** open a public issue for those.
|
provael-0.1.0/LICENSE
ADDED
|
@@ -0,0 +1,201 @@
|
|
|
1
|
+
Apache License
|
|
2
|
+
Version 2.0, January 2004
|
|
3
|
+
http://www.apache.org/licenses/
|
|
4
|
+
|
|
5
|
+
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
|
|
6
|
+
|
|
7
|
+
1. Definitions.
|
|
8
|
+
|
|
9
|
+
"License" shall mean the terms and conditions for use, reproduction,
|
|
10
|
+
and distribution as defined by Sections 1 through 9 of this document.
|
|
11
|
+
|
|
12
|
+
"Licensor" shall mean the copyright owner or entity authorized by
|
|
13
|
+
the copyright owner that is granting the License.
|
|
14
|
+
|
|
15
|
+
"Legal Entity" shall mean the union of the acting entity and all
|
|
16
|
+
other entities that control, are controlled by, or are under common
|
|
17
|
+
control with that entity. For the purposes of this definition,
|
|
18
|
+
"control" means (i) the power, direct or indirect, to cause the
|
|
19
|
+
direction or management of such entity, whether by contract or
|
|
20
|
+
otherwise, or (ii) ownership of fifty percent (50%) or more of the
|
|
21
|
+
outstanding shares, or (iii) beneficial ownership of such entity.
|
|
22
|
+
|
|
23
|
+
"You" (or "Your") shall mean an individual or Legal Entity
|
|
24
|
+
exercising permissions granted by this License.
|
|
25
|
+
|
|
26
|
+
"Source" form shall mean the preferred form for making modifications,
|
|
27
|
+
including but not limited to software source code, documentation
|
|
28
|
+
source, and configuration files.
|
|
29
|
+
|
|
30
|
+
"Object" form shall mean any form resulting from mechanical
|
|
31
|
+
transformation or translation of a Source form, including but
|
|
32
|
+
not limited to compiled object code, generated documentation,
|
|
33
|
+
and conversions to other media types.
|
|
34
|
+
|
|
35
|
+
"Work" shall mean the work of authorship, whether in Source or
|
|
36
|
+
Object form, made available under the License, as indicated by a
|
|
37
|
+
copyright notice that is included in or attached to the work
|
|
38
|
+
(an example is provided in the Appendix below).
|
|
39
|
+
|
|
40
|
+
"Derivative Works" shall mean any work, whether in Source or Object
|
|
41
|
+
form, that is based on (or derived from) the Work and for which the
|
|
42
|
+
editorial revisions, annotations, elaborations, or other modifications
|
|
43
|
+
represent, as a whole, an original work of authorship. For the purposes
|
|
44
|
+
of this License, Derivative Works shall not include works that remain
|
|
45
|
+
separable from, or merely link (or bind by name) to the interfaces of,
|
|
46
|
+
the Work and Derivative Works thereof.
|
|
47
|
+
|
|
48
|
+
"Contribution" shall mean any work of authorship, including
|
|
49
|
+
the original version of the Work and any modifications or additions
|
|
50
|
+
to that Work or Derivative Works thereof, that is intentionally
|
|
51
|
+
submitted to Licensor for inclusion in the Work by the copyright owner
|
|
52
|
+
or by an individual or Legal Entity authorized to submit on behalf of
|
|
53
|
+
the copyright owner. For the purposes of this definition, "submitted"
|
|
54
|
+
means any form of electronic, verbal, or written communication sent
|
|
55
|
+
to the Licensor or its representatives, including but not limited to
|
|
56
|
+
communication on electronic mailing lists, source code control systems,
|
|
57
|
+
and issue tracking systems that are managed by, or on behalf of, the
|
|
58
|
+
Licensor for the purpose of discussing and improving the Work, but
|
|
59
|
+
excluding communication that is conspicuously marked or otherwise
|
|
60
|
+
designated in writing by the copyright owner as "Not a Contribution."
|
|
61
|
+
|
|
62
|
+
"Contributor" shall mean Licensor and any individual or Legal Entity
|
|
63
|
+
on behalf of whom a Contribution has been received by Licensor and
|
|
64
|
+
subsequently incorporated within the Work.
|
|
65
|
+
|
|
66
|
+
2. Grant of Copyright License. Subject to the terms and conditions of
|
|
67
|
+
this License, each Contributor hereby grants to You a perpetual,
|
|
68
|
+
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
|
|
69
|
+
copyright license to reproduce, prepare Derivative Works of,
|
|
70
|
+
publicly display, publicly perform, sublicense, and distribute the
|
|
71
|
+
Work and such Derivative Works in Source or Object form.
|
|
72
|
+
|
|
73
|
+
3. Grant of Patent License. Subject to the terms and conditions of
|
|
74
|
+
this License, each Contributor hereby grants to You a perpetual,
|
|
75
|
+
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
|
|
76
|
+
(except as stated in this section) patent license to make, have made,
|
|
77
|
+
use, offer to sell, sell, import, and otherwise transfer the Work,
|
|
78
|
+
where such license applies only to those patent claims licensable
|
|
79
|
+
by such Contributor that are necessarily infringed by their
|
|
80
|
+
Contribution(s) alone or by combination of their Contribution(s)
|
|
81
|
+
with the Work to which such Contribution(s) was submitted. If You
|
|
82
|
+
institute patent litigation against any entity (including a
|
|
83
|
+
cross-claim or counterclaim in a lawsuit) alleging that the Work
|
|
84
|
+
or a Contribution incorporated within the Work constitutes direct
|
|
85
|
+
or contributory patent infringement, then any patent licenses
|
|
86
|
+
granted to You under this License for that Work shall terminate
|
|
87
|
+
as of the date such litigation is filed.
|
|
88
|
+
|
|
89
|
+
4. Redistribution. You may reproduce and distribute copies of the
|
|
90
|
+
Work or Derivative Works thereof in any medium, with or without
|
|
91
|
+
modifications, and in Source or Object form, provided that You
|
|
92
|
+
meet the following conditions:
|
|
93
|
+
|
|
94
|
+
(a) You must give any other recipients of the Work or Derivative
|
|
95
|
+
Works a copy of this License; and
|
|
96
|
+
|
|
97
|
+
(b) You must cause any modified files to carry prominent notices
|
|
98
|
+
stating that You changed the files; and
|
|
99
|
+
|
|
100
|
+
(c) You must retain, in the Source form of any Derivative Works
|
|
101
|
+
that You distribute, all copyright, patent, trademark, and
|
|
102
|
+
attribution notices from the Source form of the Work, excluding
|
|
103
|
+
those notices that do not pertain to any part of the Derivative
|
|
104
|
+
Works; and
|
|
105
|
+
|
|
106
|
+
(d) If the Work includes a "NOTICE" text file as part of its
|
|
107
|
+
distribution, then any Derivative Works that You distribute must
|
|
108
|
+
include a readable copy of the attribution notices contained
|
|
109
|
+
within such NOTICE file, excluding those notices that do not
|
|
110
|
+
pertain to any part of the Derivative Works, in at least one
|
|
111
|
+
of the following places: within a NOTICE text file distributed
|
|
112
|
+
as part of the Derivative Works; within the Source form or
|
|
113
|
+
documentation, if provided along with the Derivative Works; or,
|
|
114
|
+
within a display generated by the Derivative Works, if and
|
|
115
|
+
wherever such third-party notices normally appear. The contents
|
|
116
|
+
of the NOTICE file are for informational purposes only and do
|
|
117
|
+
not modify the License. You may add Your own attribution notices
|
|
118
|
+
within Derivative Works that You distribute, alongside or as an
|
|
119
|
+
addendum to the NOTICE text from the Work, provided that such
|
|
120
|
+
additional attribution notices cannot be construed as modifying
|
|
121
|
+
the License.
|
|
122
|
+
|
|
123
|
+
You may add Your own copyright statement to Your modifications and
|
|
124
|
+
may provide additional or different license terms and conditions
|
|
125
|
+
for use, reproduction, or distribution of Your modifications, or
|
|
126
|
+
for any such Derivative Works as a whole, provided Your use,
|
|
127
|
+
reproduction, and distribution of the Work otherwise complies with
|
|
128
|
+
the conditions stated in this License.
|
|
129
|
+
|
|
130
|
+
5. Submission of Contributions. Unless You explicitly state otherwise,
|
|
131
|
+
any Contribution intentionally submitted for inclusion in the Work
|
|
132
|
+
by You to the Licensor shall be under the terms and conditions of
|
|
133
|
+
this License, without any additional terms or conditions.
|
|
134
|
+
Notwithstanding the above, nothing herein shall supersede or modify
|
|
135
|
+
the terms of any separate license agreement you may have executed
|
|
136
|
+
with Licensor regarding such Contributions.
|
|
137
|
+
|
|
138
|
+
6. Trademarks. This License does not grant permission to use the trade
|
|
139
|
+
names, trademarks, service marks, or product names of the Licensor,
|
|
140
|
+
except as required for reasonable and customary use in describing the
|
|
141
|
+
origin of the Work and reproducing the content of the NOTICE file.
|
|
142
|
+
|
|
143
|
+
7. Disclaimer of Warranty. Unless required by applicable law or
|
|
144
|
+
agreed to in writing, Licensor provides the Work (and each
|
|
145
|
+
Contributor provides its Contributions) on an "AS IS" BASIS,
|
|
146
|
+
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
|
|
147
|
+
implied, including, without limitation, any warranties or conditions
|
|
148
|
+
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
|
|
149
|
+
PARTICULAR PURPOSE. You are solely responsible for determining the
|
|
150
|
+
appropriateness of using or redistributing the Work and assume any
|
|
151
|
+
risks associated with Your exercise of permissions under this License.
|
|
152
|
+
|
|
153
|
+
8. Limitation of Liability. In no event and under no legal theory,
|
|
154
|
+
whether in tort (including negligence), contract, or otherwise,
|
|
155
|
+
unless required by applicable law (such as deliberate and grossly
|
|
156
|
+
negligent acts) or agreed to in writing, shall any Contributor be
|
|
157
|
+
liable to You for damages, including any direct, indirect, special,
|
|
158
|
+
incidental, or consequential damages of any character arising as a
|
|
159
|
+
result of this License or out of the use or inability to use the
|
|
160
|
+
Work (including but not limited to damages for loss of goodwill,
|
|
161
|
+
work stoppage, computer failure or malfunction, or any and all
|
|
162
|
+
other commercial damages or losses), even if such Contributor
|
|
163
|
+
has been advised of the possibility of such damages.
|
|
164
|
+
|
|
165
|
+
9. Accepting Warranty or Additional Liability. While redistributing
|
|
166
|
+
the Work or Derivative Works thereof, You may choose to offer,
|
|
167
|
+
and charge a fee for, acceptance of support, warranty, indemnity,
|
|
168
|
+
or other liability obligations and/or rights consistent with this
|
|
169
|
+
License. However, in accepting such obligations, You may act only
|
|
170
|
+
on Your own behalf and on Your sole responsibility, not on behalf
|
|
171
|
+
of any other Contributor, and only if You agree to indemnify,
|
|
172
|
+
defend, and hold each Contributor harmless for any liability
|
|
173
|
+
incurred by, or claims asserted against, such Contributor by reason
|
|
174
|
+
of your accepting any such warranty or additional liability.
|
|
175
|
+
|
|
176
|
+
END OF TERMS AND CONDITIONS
|
|
177
|
+
|
|
178
|
+
APPENDIX: How to apply the Apache License to your work.
|
|
179
|
+
|
|
180
|
+
To apply the Apache License to your work, attach the following
|
|
181
|
+
boilerplate notice, with the fields enclosed by brackets "[]"
|
|
182
|
+
replaced with your own identifying information. (Don't include
|
|
183
|
+
the brackets!) The text should be enclosed in the appropriate
|
|
184
|
+
comment syntax for the file format. We also recommend that a
|
|
185
|
+
file or class name and description of purpose be included on the
|
|
186
|
+
same "printed page" as the copyright notice for easier
|
|
187
|
+
identification within third-party archives.
|
|
188
|
+
|
|
189
|
+
Copyright [yyyy] [name of copyright owner]
|
|
190
|
+
|
|
191
|
+
Licensed under the Apache License, Version 2.0 (the "License");
|
|
192
|
+
you may not use this file except in compliance with the License.
|
|
193
|
+
You may obtain a copy of the License at
|
|
194
|
+
|
|
195
|
+
http://www.apache.org/licenses/LICENSE-2.0
|
|
196
|
+
|
|
197
|
+
Unless required by applicable law or agreed to in writing, software
|
|
198
|
+
distributed under the License is distributed on an "AS IS" BASIS,
|
|
199
|
+
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
200
|
+
See the License for the specific language governing permissions and
|
|
201
|
+
limitations under the License.
|