SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2202.07261 · ECCV · 2022

Exploring the Devil in Graph Spectral Domain for 3D Point Cloud Attacks

Wei Hu, Daizong Liu, Qianjiang Hu

arXiv · PDF · Open in the Atlas

Code that ran

9 of 10 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, 0 violated it and 6 have no contract verdict; 2 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/2202.07261.json · badge: /api/ran/2202.07261.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
WoodwindHu/GSDA canonical (classifier label)
classifier basis not recorded on the link
9 ran · 1 no recorded run
FunctionStatusWhere it lives
jitter_input Ran
ran on a synthesized input
WoodwindHu/GSDA/Lib/utility.py
code served (permissive licence) · get_code("ac3b44ab95cf8165")
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 Tensor, ndim 3, dtype family float, shape (2, 3, 4) 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

torch.manual_seed(0)

# The function expects:
# - data: a tensor with shape (B, 3, N) where B is batch size, 3 is fixed, N is number of points
# - sigma: float, standard deviation for jitter (default 0.01)
# - clip: float, clipping range (default 0.05)

# Create a tensor with shape (B=2, 3, N=4)
data = torch.randn(2, 3, 4)

ARGS = {
    "data": data,
    "sigma": 0.01,
    "clip": 0.05,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
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 3, dtype family float, shape (2, 3, 4) — shape (2, 3, 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
load_pointcloud Ran
ran on a synthesized input
WoodwindHu/GSDA/distance.py
code served (permissive licence) · get_code("0e01a07c931ea6bc")
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 ndarray, ndim 2, dtype family float, shape (1024, 3) 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
import tempfile
import os

# Create a temporary CSV file with point cloud data
with tempfile.NamedTemporaryFile(mode='w', suffix='.csv', delete=False) as f:
    temp_file = f.name
    # Write 1024 points with 3 coordinates (x, y, z)
    for i in range(1024):
        x = np.random.randn()
        y = np.random.randn()
        z = np.random.randn()
        f.write(f"{x},{y},{z}\n")

ARGS = {
    'file_path': temp_file
}
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 ndarray, ndim 2, dtype family float, shape (1024, 3) — shape (1024, 3) 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
norm_l2_loss Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
WoodwindHu/GSDA/Lib/loss_utils.py
code served (permissive licence) · get_code("702f0f77f4137090")
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 Tensor, ndim 1, dtype family float, shape (2,) 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
import numpy as np

torch.manual_seed(0)

# The function norm_l2_loss takes two point cloud tensors
# and computes L2 loss by:
# 1. Computing element-wise difference
# 2. Squaring it
# 3. Summing over dimension 1 (features/coordinates)
# 4. Summing over dimension 1 again (this appears to be a mistake in the function, but we follow it)
# 
# Expected shapes: (batch_size, num_points, num_features) or similar
# The function does .sum(1).sum(1), suggesting at least 3D tensors

# Create two small point cloud tensors
# Shape: (batch_size=2, num_points=4, features=3)
adv_pc = torch.randn(2, 4, 3)
ori_pc = torch.randn(2, 4, 3)

ARGS = {
    "adv_pc": adv_pc,
    "ori_pc": ori_pc
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 1, dtype family float, shape (2,) — shape (2,) 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
outlier_removal_fn Ran
ran on a synthesized input
WoodwindHu/GSDA/defense.py
code served (permissive licence) · get_code("2bddbf49c38dd521")
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 torch
torch.manual_seed(0)

# Construct a point cloud tensor: shape (batch_size, 3, num_points)
# batch_size=2, 3 coordinates (x,y,z), num_points=4
pc = torch.randn(2, 3, 4)

# defense_type: string indicating which defense method to use
defense_type = 'outliers_variance'

# drop_num: number of points to drop (used in 'outliers_fixNum' mode)
drop_num = 1

# alpha: standard deviation multiplier for outlier threshold
alpha = 1.5

# outlier_knn: number of nearest neighbors to consider
outlier_knn = 2

ARGS = {
    'pc': pc,
    'defense_type': defense_type,
    'drop_num': drop_num,
    'alpha': alpha,
    'outlier_knn': outlier_knn
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — 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
pc_normalize Ran
ran, honoured the contract we drafted
WoodwindHu/GSDA/distance.py
pointer only (licence: MIT); the stored copy get_code serves, from aboulch/ConvPoint, is pointer only (licence: NOASSERTION) · get_code("ec413739d406e611")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from aboulch/ConvPoint, and GitHub could not classify the licence of that repository (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

point_removal_fn Ran
ran on a synthesized input
WoodwindHu/GSDA/defense.py
code served (permissive licence) · get_code("78c7047c8cc0f3d2")
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. 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 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

torch.manual_seed(0)

# Create a small point cloud tensor
# Shape: (batch_size, num_features, num_points)
# batch_size=2, num_features=3 (x,y,z), num_points=4
pc = torch.randn(2, 3, 4)

# Parameters for point_removal_fn
defense_type = 'rand_drop'
drop_num = 1
alpha = 2.0
outlier_knn = 2

ARGS = {
    'pc': pc,
    'defense_type': defense_type,
    'drop_num': drop_num,
    'alpha': alpha,
    'outlier_knn': outlier_knn
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — 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
random_drop_fn Ran
ran on a synthesized input
WoodwindHu/GSDA/defense.py
code served (permissive licence) · get_code("5a2510c05015f90f")
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. 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 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

torch.manual_seed(0)

# pc: a point cloud tensor with shape (batch, channels, num_points)
# The function accesses pc.size(2) to get num_points, clones it, and indexes dimension 2
pc = torch.randn(2, 4, 8)

# drop_num: number of points to drop (must be < num_points)
drop_num = 3

ARGS = {
    "pc": pc,
    "drop_num": drop_num,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — 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
set_bn_momentum_default Ran
ran; our contract draft was wrong, not the code
WoodwindHu/GSDA/Model/PointNetPP_ssg.py
code served (permissive licence) · get_code("0ad12b1e8408e2f9")
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 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)

# The function set_bn_momentum_default takes a bn_momentum parameter
# and returns a function that sets the momentum attribute on BatchNorm modules.
# The bn_momentum parameter should be a float value (typically between 0 and 1).

ARGS = {
    "bn_momentum": 0.1
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
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
spectral_loss Ran
ran on a synthesized input
WoodwindHu/GSDA/Lib/loss_utils.py
code served (permissive licence) · get_code("9e67731e2a1fd391")
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 Tensor, ndim 1, dtype family float, shape (2,) 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

torch.manual_seed(0)

# adv_pc: adversarial point cloud, shape (batch, num_points, features)
# ori_pc: original point cloud, shape (batch, num_points, features)
# v: transformation matrix, shape (batch, features, latent_dim)

batch_size = 2
num_points = 4
input_features = 3
latent_dim = 5

adv_pc = torch.randn(batch_size, num_points, input_features)
ori_pc = torch.randn(batch_size, num_points, input_features)
v = torch.randn(batch_size, input_features, latent_dim)

ARGS = {
    'adv_pc': adv_pc,
    'ori_pc': ori_pc,
    'v': v
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
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 1, dtype family float, shape (2,) — shape (2,) 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
array2samples_distance Not yet run
unverified (harvested, no recorded run)
WoodwindHu/GSDA/distance.py
code served (permissive licence) · get_code("947c203f90fe2111")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 distinct functions with an identity, 3 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 3 later) and 7 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.

FunctionAlso present inEarliest-datedOthers named
pc_normalize 36 other papers
11 dated earlier, 25 later
1803.05827 (2018-03)
Local Spectral Graph Convolution for Point Set Feature Learning
1904.02375 (2019-04)
ConvPoint: Continuous Convolutions for Point Cloud Processing
1904.10014 (2019-04)
Linked Dynamic Graph CNN: Learning on Point Cloud via Linking Hierarchical Features
and 33 more not named here
set_bn_momentum_default 16 other papers
4 dated earlier, 12 later
1904.09664 (2019-04)
Deep Hough Voting for 3D Object Detection in Point Clouds
1911.11236 (2019-11)
RandLA-Net: Efficient Semantic Segmentation of Large-Scale Point Clouds
2004.05679 (2020-04)
MLCVNet: Multi-Level Context VoteNet for 3D Object Detection
and 13 more not named here
norm_l2_loss 1 other paper
1 later
2508.15650 (2025-08)
Towards a 3D Transfer-based Black-box Attack via Critical Feature Guidance

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

With the maturity of depth sensors, point clouds have received increasing attention in various applications such as autonomous driving, robotics, surveillance, etc., while deep point cloud learning models have shown to be vulnerable to adversarial attacks. Existing attack methods generally add/delete points or perform point-wise perturbation over point clouds to generate adversarial examples in the data space, which may neglect the geometric characteristics of point clouds. Instead, we propose point cloud attacks from a new perspective-Graph Spectral Domain Attack (GSDA), aiming to perturb transform coefficients in the graph spectral domain that corresponds to varying certain geometric structure. In particular, we naturally represent a point cloud over a graph, and adaptively transform the coordinates of points into the graph spectral domain via graph Fourier transform (GFT) for compact representation. We then analyze the influence of different spectral bands on the geometric structure of the point cloud, based on which we propose to perturb the GFT coefficients in a learnable manner guided by an energy constraint loss function. Finally, the adversarial point cloud is generated by transforming the perturbed spectral representation back to the data domain via the inverse GFT (IGFT). Experimental results demonstrate the effectiveness of the proposed GSDA in terms of both imperceptibility and attack success rates under a variety of defense strategies. The code is available at https://github.com/WoodwindHu/GSDA.

For agents

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

get_harvested_code_for_paper("2202.07261")
get_code_for_paper("2202.07261")
have("2202.07261")

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

curl https://syntology.ai/api/ran/2202.07261.json

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

[![Syntology run record](https://syntology.ai/api/ran/2202.07261.svg)](https://syntology.ai/paper/2202.07261)

Connect an agent — have() is free.