SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2003.10596 · 2020

Adversarial Perturbations Fool Deepfake Detectors

arXiv · PDF · Open in the Atlas

Code that ran

5 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; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 5 failures of our instrument; 1 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/2003.10596.json · badge: /api/ran/2003.10596.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
ApGa/adversarial_deepfakes canonical (classifier label)
classifier basis: classifier_2025
5 ran · 1 no recorded run
FunctionStatusWhere it lives
add_noise Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ApGa/adversarial_deepfakes/classifier.py
code served (permissive licence) · get_code("24610ea98446fa84")
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) and returned Tensor, 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 torch
from torch.autograd import Variable

torch.manual_seed(0)

# Create a small input tensor with values in [0, 1] range
input_tensor = torch.rand(2, 4)

# Test with Gaussian noise
ARGS = {
    'input': input_tensor,
    'noise_type': 'Gaussian',
    'noise1': 0.0,
    'noise2': 0.1
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, 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
bim Ran
ran; our contract draft was wrong, not the code
ApGa/adversarial_deepfakes/adv_examples.py
code served (permissive licence) · get_code("c54694cd7ab40f55")
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) and returned function. 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 torch
import torch.nn as nn

torch.manual_seed(0)

# Define a simple model that takes an image and outputs logits
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.fc = nn.Linear(4, 3)  # Input dim 4, output dim 3 (3 classes)
    
    def forward(self, x):
        return self.fc(x)

# Create model instance
model = SimpleModel()

# Define a loss function (CrossEntropyLoss for classification)
loss = nn.CrossEntropyLoss()

# Define epsilon for perturbation
eps = 0.1

ARGS = {
    "model": model,
    "loss": loss,
    "eps": eps,
    "iters": 4,
    "softmax": False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
fgsm Ran
ran; our contract draft was wrong, not the code
ApGa/adversarial_deepfakes/adv_examples.py
code served (permissive licence) · get_code("969781f75548142e")
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) and returned function. 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 torch
import torch.nn as nn

torch.manual_seed(0)

# Build a minimal model that takes an image tensor and outputs logits
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)  # 4 input features, 3 output classes
    
    def forward(self, x):
        return self.linear(x)

model = SimpleModel()

# Create a loss function - CrossEntropyLoss for classification
loss = nn.CrossEntropyLoss()

# Small epsilon for perturbation
eps = 0.1

ARGS = {
    "model": model,
    "loss": loss,
    "eps": eps,
    "softmax": False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
ifgsm Ran
ran; our contract draft was wrong, not the code
ApGa/adversarial_deepfakes/adv_examples.py
code served (permissive licence) · get_code("c1a53b05fe8519cd")
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) and returned function. 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 torch
import torch.nn as nn

torch.manual_seed(0)

# Define a simple model that takes an image and outputs logits
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.fc = nn.Linear(4, 3)  # Input dim 4, output dim 3 (3 classes)
    
    def forward(self, x):
        # Flatten if needed and pass through linear layer
        return self.fc(x)

# Create model instance
model = SimpleModel()

# Define a loss function - using CrossEntropyLoss which expects logits and class indices
loss = nn.CrossEntropyLoss()

# Define epsilon for perturbation
eps = 0.1

# Define number of iterations
iters = 4

# Define softmax flag
softmax = False

ARGS = {
    "model": model,
    "loss": loss,
    "eps": eps,
    "iters": iters,
    "softmax": softmax
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
lipshitz_regularization Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ApGa/adversarial_deepfakes/classifier.py
code served (permissive licence) · get_code("24f2b9a08b2cebda")
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) and returned Tensor, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 torch
import torch.nn as nn

torch.manual_seed(0)

# Create a minimal model that takes images and returns output
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.conv = nn.Conv2d(3, 8, kernel_size=1)
        self.fc = nn.Linear(8 * 4 * 4, 2)
    
    def forward(self, x):
        # x shape: (batch, 3, 4, 4)
        x = self.conv(x)  # (batch, 8, 4, 4)
        x = x.view(x.size(0), -1)  # (batch, 128)
        x = self.fc(x)  # (batch, 2)
        return x

# Create model instance
model = SimpleModel()

# Create images tensor with requires_grad=True for gradient computation
# Shape: (batch=2, channels=3, height=4, width=4)
images = torch.randn(2, 3, 4, 4, requires_grad=True)

# Define all arguments
ARGS = {
    'images': images,
    'model': model,
    'z': 2,
    'train': True,
    'psi': 1000,
    'use_softmax': True,
    'num_classes': 2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, 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
train_model Not yet run
unverified (harvested, no recorded run)
ApGa/adversarial_deepfakes/classifier.py
code served (permissive licence) · get_code("0e2498182c6b0d20")
How it ran

no invocation recorded.

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

This work uses adversarial perturbations to enhance deepfake images and fool common deepfake detectors. We created adversarial perturbations using the Fast Gradient Sign Method and the Carlini and Wagner L2 norm attack in both blackbox and whitebox settings. Detectors achieved over 95% accuracy on unperturbed deepfakes, but less than 27% accuracy on perturbed deepfakes. We also explore two improvements to deepfake detectors: (i) Lipschitz regularization, and (ii) Deep Image Prior (DIP). Lipschitz regularization constrains the gradient of the detector with respect to the input in order to increase robustness to input perturbations. The DIP defense removes perturbations using generative convolutional neural networks in an unsupervised manner. Regularization improved the detection of perturbed deepfakes on average, including a 10% accuracy boost in the blackbox case. The DIP defense achieved 95% accuracy on perturbed deepfakes that fooled the original detector, while retaining 98% accuracy in other cases on a 100 image subsample.

For agents

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

get_harvested_code_for_paper("2003.10596")
get_code_for_paper("2003.10596")
have("2003.10596")

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

curl https://syntology.ai/api/ran/2003.10596.json

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

[![Syntology run record](https://syntology.ai/api/ran/2003.10596.svg)](https://syntology.ai/paper/2003.10596)

Connect an agent — have() is free.