AGENT SKILL

privacy-tests

Routes TensorFlow Privacy users who want membership inference analysis, privacy reports, membership inference callbacks, or secret-sharer exposure checks.

License 许可
Apache-2.0
最近更新
2026-09-03
Stars
★ 351
Forks
27

01 前言 Front Matter

name
privacy-tests
license
Apache 2.0
metadata
{ "disco-role": "operating" }
description
Routes TensorFlow Privacy users who want membership inference analysis, privacy reports, membership inference callbacks, or secret-sharer exposure checks.
disable-model-invocation
true

02 资源文件 4

03 SKILL.md 正文

Privacy tests

Use this sub-skill when the user wants to measure memorization or privacy risk from model outputs.

Trigger phrases

  • "membership inference"
  • "privacy report"
  • "attack results"
  • "epsilon lower bound"
  • "privacy callback"
  • "secret sharer"
  • "exposure"
  • "memorization"

What this sub-skill covers

  • AttackInputData, SlicingSpec, and AttackType
  • run_attacks(), run_membership_probability_analysis(), and run_attack_on_keras_model()
  • MembershipInferenceCallback and the keras_evaluation callback path
  • AttackResults, privacy-report metadata, and pandas summaries
  • privacy_report plotting/report helpers
  • secret-sharer secret generation and exposure computation

What it does not cover

  • DP training or optimizer selection -> ../training/
  • privacy budget and epsilon/noise search -> ../privacy-accounting/
  • DPQuery internals -> ../queries/
  • fast clipping internals -> ../fast-clipping/

Read this before you act

  • references/api-reference.md for attack inputs, outputs, and the bundled smoke pattern.
  • references/troubleshooting.md for tiny-data, shape, and dependency failures.
  • ../../references/install-and-scope.md for the minimum CPU runtime.

Typical workflow

  1. Decide whether the user wants membership inference or secret-sharer exposure analysis.
  2. Build the smallest valid AttackInputData or secret-sharer fixture.
  3. Use the bundled smoke helper before trying a larger analysis.
  4. If the user is asking for training guidance rather than analysis, route to training.

Bundled helpers

  • scripts/tiny_mia_smoke.py runs a threshold membership-inference attack on synthetic losses.
  • scripts/tiny_secret_sharer_smoke.py generates a toy secret set and computes exposure on synthetic perplexities.

04 同仓库更多 Skills

VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-triton-opt

Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-solve

END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-skill

PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-pattern-fusion

Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-orchestrate

Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-feedback

Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).

★ 351