SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2109.08266 · 2021

Hard to Forget: Poisoning Attacks on Certified Machine Unlearning

arXiv · PDF · Open in the Atlas

Code that ran

2 of 11 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 2 have no contract verdict; 0 failures of our instrument; 9 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2109.08266.json · badge: /api/ran/2109.08266.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
ngmarchant/attack-unlearning canonical (classifier label)
classifier basis: census
2 ran · 9 no recorded run
FunctionStatusWhere it lives
l1_proj Ran
ran on a synthesized input
ngmarchant/attack-unlearning/projections.py
code served (permissive licence) · get_code("fd3e40a8f8b31a66")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ArrayImpl, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import jax.numpy as jnp
import jax

# Seed for reproducibility
jax.config.update("jax_platform_name", "cpu")

# Create small test arrays
# eps: float parameter for L1 ball scaling
eps = 1.5

# x: input array (small, 2D for clarity)
x = jnp.array([[1.2, -0.5, 2.1, -1.3],
               [0.8, 1.5, -0.9, 0.3]], dtype=jnp.float32)

# x_ref: optional reference point (center of L1 ball)
# Use a small array with same shape as x
x_ref = jnp.array([[0.1, 0.0, 0.2, -0.1],
                   [0.0, 0.1, -0.05, 0.0]], dtype=jnp.float32)

# axis: optional axis parameter (set to None for full array projection)
axis = None

ARGS = {
    "eps": eps,
    "x": x,
    "x_ref": x_ref,
    "axis": axis,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
tree_zeros_like Ran
ran on a synthesized input
ngmarchant/attack-unlearning/util.py
code served (permissive licence) · get_code("503701aad49b036e")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import jax
import jax.numpy as jnp
from jax.flatten_util import ravel_pytree

# Create a simple pytree structure
# The function expects a pytree (nested structure of arrays/values)
# We'll create a dictionary with some JAX arrays

tree = {
    'layer1': jnp.array([1.0, 2.0, 3.0, 4.0]),
    'layer2': {
        'weights': jnp.array([[1.0, 2.0], [3.0, 4.0], [5.0, 6.0]]),
        'bias': jnp.array([0.5, 1.5])
    },
    'scalar': jnp.array(42.0)
}

ARGS = {
    'tree': tree
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
fashion_mnist Not yet run
unverified (harvested, no recorded run)
ngmarchant/attack-unlearning/datasets.py
code served (permissive licence) · get_code("8a5e359aa5894852")
How it ran

no invocation recorded.

lbfgs_minimize Not yet run
unverified (harvested, no recorded run)
ngmarchant/attack-unlearning/lbfgs_minimize.py
code served (permissive licence) · get_code("16544f7487331b0e")
How it ran

no invocation recorded.

linf_proj Not yet run
unverified (harvested, no recorded run)
ngmarchant/attack-unlearning/projections.py
code served (permissive licence) · get_code("ac0824e8d3c3c2b8")
How it ran

no invocation recorded.

mnist Not yet run
unverified (harvested, no recorded run)
ngmarchant/attack-unlearning/datasets.py
code served (permissive licence) · get_code("2d79a5710c3e0b38")
How it ran

no invocation recorded.

mnist_binary Not yet run
unverified (harvested, no recorded run)
ngmarchant/attack-unlearning/datasets.py
code served (permissive licence) · get_code("9fa62c93e01e3e6b")
How it ran

no invocation recorded.

normalize Not yet run
unverified (harvested, no recorded run)
ngmarchant/attack-unlearning/init_experiment.py
code served (permissive licence) · get_code("efd91a4d57b99306")
How it ran

no invocation recorded.

proj_grad_descent Not yet run
unverified (harvested, no recorded run)
ngmarchant/attack-unlearning/grad_descent.py
code served (permissive licence) · get_code("3936f0ddb76eb43a")
How it ran

no invocation recorded.

register_pytree_node_dataclass Not yet run
unverified (harvested, no recorded run)
ngmarchant/attack-unlearning/util.py
code served (permissive licence) · get_code("303a2ef6d3d8fe00")
How it ran

no invocation recorded.

soft_thresh Not yet run
unverified (harvested, no recorded run)
ngmarchant/attack-unlearning/projections.py
code served (permissive licence) · get_code("1e6c97b7639df426")
How it ran

no invocation recorded.

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 11 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

The right to erasure requires removal of a user's information from data held by organizations, with rigorous interpretations extending to downstream products such as learned models. Retraining from scratch with the particular user's data omitted fully removes its influence on the resulting model, but comes with a high computational cost. Machine "unlearning" mitigates the cost incurred by full retraining: instead, models are updated incrementally, possibly only requiring retraining when approximation errors accumulate. Rapid progress has been made towards privacy guarantees on the indistinguishability of unlearned and retrained models, but current formalisms do not place practical bounds on computation. In this paper we demonstrate how an attacker can exploit this oversight, highlighting a novel attack surface introduced by machine unlearning. We consider an attacker aiming to increase the computational cost of data removal. We derive and empirically investigate a poisoning attack on certified machine unlearning where strategically designed training data triggers complete retraining when removed.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2109.08266")
get_code_for_paper("2109.08266")
have("2109.08266")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2109.08266.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2109.08266.svg)](https://syntology.ai/paper/2109.08266)

Connect an agent — have() is free.