SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2510.15083 · ICLR · 2025

SMOTE and Mirrors: Exposing Privacy Leakage from Synthetic Minority Oversampling *

Georgi Ganev, Emiliano De Cristofaro, Reza Nazari, Rees Davison, Amir Dizche, Xinmin Wu, Ralph Abbey, Jorge Silva, Ucl

arXiv · PDF · Open in the Atlas

Code that ran

6 of 6 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; 1 honoured the contract we drafted, 1 violated it and 0 have no contract verdict; 4 failures of our instrument; 0 no recorded run), measured 2026-11-22; 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/2510.15083.json · badge: /api/ran/2510.15083.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
sascommunities/smote-mirrors canonical (classifier label)
classifier basis: census
6 ran · 0 no recorded run
FunctionStatusWhere it lives
check_collinear Ran
ran, violated the contract we drafted
sascommunities/smote-mirrors/code/attacks/utils.py
code served (permissive licence) · get_code("e0cdc713f55f8fb4")
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 bool, ndim 0, dtype family float, shape () 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 numpy as np

# Test case 1: collinear points (should return True)
a1 = np.array([0.0, 0.0, 0.0])
b1 = np.array([1.0, 1.0, 1.0])
c1 = np.array([2.0, 2.0, 2.0])

# Test case 2: non-collinear points (should return False)
a2 = np.array([0.0, 0.0, 0.0])
b2 = np.array([1.0, 0.0, 0.0])
c2 = np.array([0.0, 1.0, 0.0])

ARGS = {
    "a": a1,
    "b": b1,
    "c": c1,
    "line_eps": 1e-12
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, ndim 0, dtype family float, shape () — shape () 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
check_midpoint Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
sascommunities/smote-mirrors/code/attacks/utils.py
code served (permissive licence) · get_code("4c7c66ddd341db5e")
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 bool, ndim 0, dtype family float, shape () 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 numpy as np

# Analyze the function:
# - Takes 4 parameters: a, b, c (appear to be numpy arrays/vectors), and line_eps (float)
# - Uses np.dot() on arrays, so a, b, c should be numpy arrays
# - Computes ab = b - a and ac = c - a, so they need compatible shapes
# - Returns a boolean checking if proj is in range (-line_eps, 1 + line_eps)

# Create small numpy vectors
a = np.array([0.0, 0.0, 0.0, 0.0])
b = np.array([1.0, 0.0, 0.0, 0.0])
c = np.array([0.5, 0.1, 0.0, 0.0])

line_eps = 1e-12

ARGS = {
    "a": a,
    "b": b,
    "c": c,
    "line_eps": line_eps
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, ndim 0, dtype family float, shape () — shape () 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
featurize_X Ran
ran, honoured the contract we drafted
sascommunities/smote-mirrors/code/attacks/membership_attack.py
code served (permissive licence) · get_code("607ef5c1ac937dbb")
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 list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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 numpy as np

# Create a small 2D numpy array with shape (5, 3)
# This will be used as input X to featurize_X
X = np.array([
    [1.0, 2.0, 3.0],
    [2.0, 3.0, 4.0],
    [3.0, 4.0, 5.0],
    [4.0, 5.0, 6.0],
    [5.0, 6.0, 7.0]
], dtype=np.float64)

ARGS = {
    "X": X
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
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
find_intersection_point Ran
ran; our contract draft was wrong, not the code
sascommunities/smote-mirrors/code/attacks/utils.py
code served (permissive licence) · get_code("54e767bd9df0a50d")
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 tuple. 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 numpy as np

# Construct valid arguments for find_intersection_point
# The function expects two sets of points and an optional line_eps parameter

# Create points_i: a small array of points (shape: n_points x dimensions)
# These points will be used to compute a line
points_i = np.array([
    [1.0, 2.0, 3.0],
    [2.0, 3.0, 4.0],
    [3.0, 4.0, 5.0]
], dtype=np.float64)

# Create points_j: another set of points representing a different line
# Construct them so they roughly intersect with points_i's line
points_j = np.array([
    [1.5, 2.5, 3.5],
    [2.5, 3.5, 4.5]
], dtype=np.float64)

# line_eps parameter for numerical tolerance
line_eps = 1e-12

ARGS = {
    "points_i": points_i,
    "points_j": points_j,
    "line_eps": line_eps
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is 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
imbalance_data Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
sascommunities/smote-mirrors/code/datasets/utils.py
code served (permissive licence) · get_code("020be57eeea6712b")
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 DataFrame. 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 pandas as pd
import numpy as np

# Create a small imbalanced dataframe for testing
np.random.seed(42)

# Create sample data with imbalanced classes
n_majority = 100
n_minority = 4

# Majority class (label 0)
majority_data = pd.DataFrame({
    'feature1': np.random.randn(n_majority),
    'feature2': np.random.randn(n_majority),
    'feature3': np.random.randn(n_majority),
    'class': 0
})

# Minority class (label 1)
minority_data = pd.DataFrame({
    'feature1': np.random.randn(n_minority),
    'feature2': np.random.randn(n_minority),
    'feature3': np.random.randn(n_minority),
    'class': 1
})

df_higgs = pd.concat([majority_data, minority_data], ignore_index=True)

# Create MiniBooNE style dataframe
df_minibooNE = pd.DataFrame({
    'feature1': np.random.randn(104),
    'feature2': np.random.randn(104),
    'signal': [b'False'] * 100 + [b'True'] * 4
})

# Create cardio style dataframe
df_cardio = pd.DataFrame({
    'feature1': np.random.randn(104),
    'feature2': np.random.randn(104),
    'cardio': [0] * 100 + [1] * 4
})

# Create churn style dataframe
df_churn = pd.DataFrame({
    'feature1': np.random.randn(104),
    'feature2': np.random.randn(104),
    'Exited': [0] * 100 + [1] * 4
})

ARGS = {
    'data_name': 'higgs',
    'df': df_higgs,
    'desired_ratio': 26
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 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
prepare_data Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
sascommunities/smote-mirrors/code/datasets/utils.py
code served (permissive licence) · get_code("3d7f429ad3a6970b")
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 tuple. 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 numpy as np
import pandas as pd
from sklearn.preprocessing import MinMaxScaler, OrdinalEncoder
from sklearn.compose import ColumnTransformer

# Create sample data
np.random.seed(0)

# Create X as a DataFrame with proper column names and types
X = pd.DataFrame({
    'age': np.random.randint(30, 65, 2),
    'gender': pd.Categorical(['M', 'F']),
    'height': np.random.randint(150, 200, 2),
    'weight': np.random.uniform(50, 100, 2),
    'ap_hi': np.random.randint(100, 180, 2),
    'ap_lo': np.random.randint(60, 120, 2),
    'cholesterol': pd.Categorical(['0', '1']),
    'gluc': pd.Categorical(['0', '1']),
    'smoke': pd.Categorical(['0', '1']),
    'alco': pd.Categorical(['0', '1']),
    'active': pd.Categorical(['0', '1']),
})

# Create y with values that will map to 0 and 1
y = np.array([-1, 1])

# Get the dtypes dict for cardio dataset
sample_dtypes = {'age': int, 'gender': 'category', 'height': int, 'weight': float, 
                 'ap_hi': int, 'ap_lo': int, 'cholesterol': 'category', 'gluc': 'category', 
                 'smoke': 'category', 'alco': 'category', 'active': 'category', 'cardio': 'category'}

ARGS = {
    'X': X,
    'y': y,
    'dtypes': sample_dtypes
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas, sklearn — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is 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

Also present in other papers' code

Of 6 samples lifted from this paper's linked repositories, 6 carry a function identity and 0 do not. Of the 6 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 6 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 Synthetic Minority Over-sampling Technique (SMOTE) is one of the most widely used methods for addressing class imbalance and generating synthetic data. Despite its popularity, little attention has been paid to its privacy implications; yet, it is used in the wild in many privacy-sensitive applications. In this work, we conduct the first systematic study of privacy leakage in SMOTE: we begin by showing that prevailing evaluation practices, i.e., naive distinguishing and distance-to-closest-record metrics, completely fail to detect any leakage and that membership inference attacks (MIAs) can be instantiated with high accuracy. Then, by exploiting SMOTE's geometric properties, we build two novel attacks with very limited assumptions: DistinSMOTE, which perfectly distinguishes real from synthetic records in augmented datasets, and ReconSMOTE, which reconstructs real minority records from synthetic datasets with perfect precision and recall approaching one under realistic imbalance ratios. We also provide theoretical guarantees for both attacks. Experiments on eight standard imbalanced datasets confirm the practicality and effectiveness of these attacks. Overall, our work reveals that SMOTE is inherently non-private and disproportionately exposes minority records, highlighting the need to reconsider its use in privacy-sensitive applications and as a baseline for assessing the privacy of modern generative models.

For agents

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

get_harvested_code_for_paper("2510.15083")
get_code_for_paper("2510.15083")
have("2510.15083")

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

curl https://syntology.ai/api/ran/2510.15083.json

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

[![Syntology run record](https://syntology.ai/api/ran/2510.15083.svg)](https://syntology.ai/paper/2510.15083)

Connect an agent — have() is free.