Pith. sign in

REVIEW 2 major objections 4 minor 37 references

Once a MAN: Towards Multi-Target Attack via Learning Multi-Target Adversarial Network Once

T0 review · 2 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

Pith's one-line read One conditional network can generate adversarial examples for any target class of a classifier, the paper claims, replacing thousands of per-class generators with a single training run.

desk verdict A useful multi-target attack extension whose headline numbers are inflated by counting the ground-truth label as a valid target; the central claim survives but needs a cleaner evaluation. read the letter →

arxiv 1908.05185 v1 pith:VEO62IFJ submitted 2019-08-14 cs.CV cs.CRcs.LG

classification cs.CVcs.CRcs.LG
keywords adversarialattackmulti-targetgeneration-basedconditionalgenerationtransferabilitytrainingimageclassificationlabelconditioning
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that a single generation network, conditioned on a target class label, can produce adversarial examples for every class of a classifier at runtime, eliminating the need for one trained generator per target class. It introduces the Multi-target Adversarial Network (MAN), with two variants that differ in how the label is fused into image features, and trains it with a classification loss through the attacked model plus a reconstruction loss. The authors report that this one model matches or beats single-target generators on white-box attack success, transfers better to unseen models, and supplies diverse adversarial samples that improve adversarial-training robustness. If true, the practical cost of targeted attack and of generating training data for defenses drops by orders of magnitude.

What carries the argument

The central object is the conditional adversarial generator: an encoder-decoder whose image-feature stream is fused with a label-embedding stream. In MANc the one-hot target label is expanded spatially and concatenated with the image features before a convolution; in MANr the label goes through a two-layer MLP with sigmoid activation and multiplies every channel of the image feature map, a squeeze-and-excitation-style recalibration. The target label is sampled at random per training image, so one forward pass of the same model can target any of the $K$ classes, and the frozen attacked classifier acts as the loss oracle that injects target information into the perturbation.

What would settle it

For a fixed trained MAN, evaluate attack success separately for each of the $K$ target classes on a held-out set: if any class's success rate drops far below the reported average, or if the label embedding $t'$ is nearly constant across different targets, then the single-model multi-target claim fails.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central claim is that target-conditional adversarial generation can be made multi-target: a single network $F_\theta(x,t)$ trained once with random target labels learns to map any input image $x$ and any target class $t$ to an adversarial image $x^*$ such that $\arg\max_i H_i(F_\theta(x,t)) = t$ for a fixed attacked classifier $H$. The paper demonstrates this with two fusion designs, one that concatenates the one-hot label map to the image features (MANc) and one that recalibrates image feature channels by multiplying them with a sigmoid-activated MLP label embedding (MANr), optimized by cross-entropy classification loss through the frozen attacked model plus an $L_2$ reconstruction loss with a distance cap. It reports attack success above 98% in white-box settings on ImageNet and CIFAR-10, stronger black-box transfer than the single-target baselines, and improved robustness when the generated samples are used to fine-tune the target classifier. The paper also reports that the multi-target model shows better transferability than a single-target model trained for the same number of iterations, attributing this to competition between target labels promoting more generalizable features.

Load-bearing premise

The load-bearing premise is that a single shared encoder-decoder, guided only by a random-target training objective, can learn to produce distinct, reliable adversarial perturbations for every one of the $K$ target classes at once without forgetting, mode collapse, or interference among classes; the paper offers empirical support but no capacity or interference analysis.

Editorial extensions

If this is right

  • Targeted attacks against classifiers with hundreds or thousands of classes no longer need one generator per class; a single MAN model covers all classes after one training run.
  • The reported black-box transfer gains imply that MAN samples generalize across architectures better than single-target generator baselines, which matters when the attacker does not know the victim model.
  • Because MAN produces diverse adversarial samples with random targets at feed-forward speed, adversarial training can be scaled to many target classes at a fraction of the training and storage cost of per-class generators.
  • The same conditional-generation idea extends beyond classification to any differentiable victim model that can provide a loss signal, though the paper only demonstrates image classifiers.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper leaves implicit that the conditional design also benefits the defense side: one MAN can generate a far larger and more diverse set of adversarial examples for data augmentation than a handful of per-class models, and the reported comparison of training with one, five, and ten per-class generators suggests diversity itself, not just target count, drives robustness gains.
  • A testable extension is to report per-target attack success separately: the paper's averages over random targets could hide tail classes where the shared model underperforms, and a per-class breakdown would show whether the label branch truly conditions the output or merely biases toward easy classes.
  • If the capacity-interference premise holds, the approach should scale to much larger label spaces, such as fine-grained datasets with thousands of classes, and to conditional attacks on other tasks such as detection or retrieval; if it fails there, a hybrid of shared features with class-specific heads may be needed.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. The paper proposes Multi-target Adversarial Network (MAN), a conditional adversarial-sample generator that takes an image x and a target label t as input and outputs a perturbed image x* via an encoder-decoder architecture. Two integration variants are presented: MANc, which concatenates image features with a one-hot label map, and MANr, which re-calibrates image feature channels through an MLP-based label encoder. The network is trained with a cross-entropy classification loss on the attacked model plus an L2 reconstruction loss, and can be used with random target labels to produce attacks against any class with a single model. Experiments cover white-box and black-box single-target attack on ImageNet and CIFAR-10, multi-target attack, use of MAN-generated samples for adversarial training, and ablations on loss weight, perturbation budget, transferability, and model ensembling. The paper claims that MAN is the first multi-target adversarial network and that it achieves competitive or better attack success and transferability than ATN and GAP, while requiring far fewer trained models to cover all classes.

Significance. If the central claim is supported, MAN would be a useful efficiency advance over training one generator per target class, and the transferability and adversarial-training results would be of interest to the adversarial ML community. The single-target comparisons in Tables 1 and 2 are internally consistent, the parameter-efficiency analysis in Table 5 is a clear strength, and the two conditional-integration variants are simple and clearly described. However, the multi-target evaluation on CIFAR-10 assigns all ten labels to every sample, including the ground-truth label, and reports only aggregate success rates; this directly affects the paper's headline claim that a single model can attack any category at the reported success rate. The manuscript is well organized and the method is conceptually reproducible, though no code is released.

major comments (2)
  1. [Section 4.1, Table 4] The CIFAR-10 multi-target evaluation assigns all ten labels to each sample, which includes the ground-truth label y. The problem definition in Section 3 explicitly requires t != y, and for t = y the classification loss in Eq. (2) is already satisfied by the original image while the reconstruction loss encourages x* approximately equal to x, so such successes are trivial. Because Table 4 reports only aggregate success over the ten targets, the reported white-box rate of 99.50% and black-box rates of 89.74-95.47% can be inflated by the trivially successful t = y cases and can hide substantially lower per-target rates for t != y. The central claim that MAN can attack any category at the reported rate is therefore not yet established. Please rerun the multi-target evaluation excluding t = y, and likewise ensure that the ImageNet random labels exclude the ground-truth label, and report per-target success rates or a source-by-target confusion matrix.
  2. [Section 4.1, training protocol, and Eq. (2)] The multi-target training phase assigns a random target label to each training image. If targets are drawn uniformly without excluding y, then on CIFAR-10 roughly 10% of training pairs have t = y, and for those pairs the classification term in Eq. (2) is near-zero while the reconstruction term dominates, producing reconstruction-only gradients with no attack signal. This is exactly the label-dilution mechanism that the shared-encoder design must overcome, and the current aggregate results do not show whether the model has truly learned all K distinct target maps. Please state whether t = y was excluded during training; if it was not, retrain or re-evaluate with y excluded and report how the aggregate multi-target numbers change. The same caveat also applies to the adversarial-training experiments in Section 4.2, where MAN-generated samples with random target labels are used to fine-tune the attacked model.
minor comments (4)
  1. [Section 3.1, Eq. (1)] The symbol t is used both for the target label index and for the one-hot vector in R^K, which makes the dimensions of W1 in Eq. (1) ambiguous; please use separate notation, e.g., e_t, for the one-hot vector.
  2. [Section 4.1, Table 3] The text says that for ImageNet the authors 'randomly assign ten labels for each sample,' but it does not state whether the ten labels are distinct or whether the ground-truth label is excluded; please specify.
  3. [Section 4.3, Figure 4(a)] The x-axis labels in Figure 4(a) are missing the symbol alpha (e.g., they should read alpha=600, alpha=800, and so on), and the caption should state that the results are on CIFAR-10 against the pretrained VGG16 model.
  4. [References] Reference [13] is cited as 'Jie et al.' in the text, but the correct author list is Hu et al.; please correct this citation.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the multi-target attack network is an empirical system validated by held-out attack rates and external black-box models, with no prediction that reduces to a fitted input or self-citation.

full rationale

MAN is an empirical construction, not a derivation. The objective in Eq. (2) combines a cross-entropy classification loss on the attacked model's output with an L2 reconstruction loss; success is measured by feeding generated samples through pretrained VGG/ResNet models and by transfer to untouched architectures. No parameter is fitted to the reported success rates, no uniqueness theorem is imported from the authors' prior work, and the cited baselines (ATN, GAP, SENet) are external works. The only in-scope concern is evaluation protocol: the CIFAR-10 multi-target test assigns all ten labels including the ground-truth class, so the aggregate rate includes a trivial target; however, this is an experimental validity issue, not circularity, because the central claim is not derived from this metric by construction and the white-box and black-box numbers remain externally falsifiable. The paper therefore scores 0.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical or theoretical entities. Its central claim is empirical and rests on standard white-box attack assumptions plus a capacity assumption for the shared multi-target generator. The free parameters are loss and perturbation hyperparameters that affect reported rates but are not fitted to reproduce a specific external result.

free parameters (2)
  • alpha (loss weight) = 100 on ImageNet, 800 on CIFAR10
    Balances classification and reconstruction losses in Eq. (2); the ablation in Fig. 4(a) shows attack success varies materially with this choice.
  • epsilon perturbation budget = 10*sqrt(N), where N=C*H*W, in main tables
    Selected by hand as the allowed L2 perturbation; all generated samples are rescaled to this bound at test time, and ablation Fig. 4(b) shows performance peaks around this value.
assumptions (4)
  • domain assumption The attacked model H is differentiable with respect to its input, so classification loss gradients can train the generator.
    Eq. (2) and Section 3.1 rely on backpropagation through the fixed attacked model, a standard white-box attack assumption.
  • ad hoc to paper A one-hot label vector combined with image features provides sufficient conditioning for target-specific perturbations.
    Section 3.1 defines MANc and MANr with this conditioning; no proof or capacity analysis is provided.
  • domain assumption Pixel-space L2 distance and test-time rescaling to a fixed L2 budget preserve attack validity and visual similarity.
    Used in Section 3.2 and the evaluation protocol in Section 4.1; common in the literature but not validated against perceptual metrics at the claimed perturbation strengths.
  • domain assumption Standard cross-entropy loss on the target classifier is a good proxy for attack success.
    Section 3.2 adopts cross-entropy without comparing alternative classification losses; the ablation only varies alpha.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Once a MAN: Towards Multi-Target Attack via Learning Multi-Target Adversarial Network Once." pith.science (2026). https://pith.science/paper/VEO62IFJ

@misc{pith2026190805185,
  author       = {Pith},
  title        = {Pith review of: Once a MAN: Towards Multi-Target Attack via Learning Multi-Target Adversarial Network Once},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VEO62IFJ}},
  note         = {Machine review of arXiv:1908.05185}
}
read the original abstract

Modern deep neural networks are often vulnerable to adversarial samples. Based on the first optimization-based attacking method, many following methods are proposed to improve the attacking performance and speed. Recently, generation-based methods have received much attention since they directly use feed-forward networks to generate the adversarial samples, which avoid the time-consuming iterative attacking procedure in optimization-based and gradient-based methods. However, current generation-based methods are only able to attack one specific target (category) within one model, thus making them not applicable to real classification systems that often have hundreds/thousands of categories. In this paper, we propose the first Multi-target Adversarial Network (MAN), which can generate multi-target adversarial samples with a single model. By incorporating the specified category information into the intermediate features, it can attack any category of the target classification model during runtime. Experiments show that the proposed MAN can produce stronger attack results and also have better transferability than previous state-of-the-art methods in both multi-target attack task and single-target attack task. We further use the adversarial samples generated by our MAN to improve the robustness of the classification model. It can also achieve better classification accuracy than other methods when attacked by various methods.

Figures

Figures reproduced from arXiv: 1908.05185 by the authors.

Figure 1
Figure 1. The comparison between our MAN and previous target [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The structure of Multi-Target Adversarial Network (MAN), (a) shows the overall architecture, (b) and (c) show two different [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Adversarial samples to different target with [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: (a) The relation between weight factor α and attack suc￾cess rate with different  on the CIFAR10 to a pretrained VGG16 model. (b) The attack accuracy under different threshold . VGG16 VGG19 Res32 Res14 Res32 Single-target 58.71 75.69 99.70∗ 90.51 Multi-target 67.69 8…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 11 canonical work pages

  1. [1]

    Adversarial transformation networks: Learning to generate adversarial examples

    Shumeet Baluja and Ian Fischer. Adversarial transformation networks: Learning to generate adversarial examples. arXiv preprint arXiv:1703.09387, 2017

  2. [2]

    Towards evaluating the robustness of neural networks

    Nicholas Carlini and David Wagner. Towards evaluating the robustness of neural networks. In 2017 IEEE Symposium on Security and Privacy (SP), pages 39–57. IEEE, 2017

  3. [3]

    Kounavis, and Duen Horng Chau

    Nilaksh Das, Madhuri Shanbhogue, Shang-Tse Chen, Fred Hohman, Li Chen, Michael E. Kounavis, and Duen Horng Chau. Keeping the Bad Guys Out: Protecting and Vaccinat- ing Deep Learning with JPEG Compression. arXiv e-prints, page arXiv:1705.02900, May 2017

  4. [4]

    Imagenet: A large-scale hierarchical im- age database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical im- age database. In Computer Vision and Pattern Recognition,

  5. [5]

    Boosting adversarial attacks with momentum

    Yinpeng Dong, Fangzhou Liao, Tianyu Pang, Hang Su, Xi- aolin Hu, Jianguo Li, and Jun Zhu. Boosting adversarial attacks with momentum. arXiv preprint arXiv:1710.06081, 2017

  6. [6]

    Practical methods of optimization

    Roger Fletcher. Practical methods of optimization . John Wiley & Sons, 2013

  7. [7]

    Convolutional sequence to sequence learning

    Jonas Gehring, Michael Auli, David Grangier, Denis Yarats, and Yann N Dauphin. Convolutional sequence to sequence learning. arXiv preprint arXiv:1705.03122, 2017

  8. [8]

    Fast r-cnn

    Ross Girshick. Fast r-cnn. In Proceedings of the IEEE inter- national conference on computer vision , pages 1440–1448, 2015

Show all 37 references
  1. [9]

    Explaining and harnessing adversarial examples

    Ian Goodfellow, Jonathon Shlens, and Christian Szegedy. Explaining and harnessing adversarial examples. In Inter- national Conference on Learning Representations, 2015

  2. [10]

    Goodfellow, Jonathon Shlens, and Christian Szegedy

    Ian J. Goodfellow, Jonathon Shlens, and Christian Szegedy. Explaining and Harnessing Adversarial Examples. arXiv e- prints, page arXiv:1412.6572, Dec 2014

  3. [11]

    Towards Deep Neural Net- work Architectures Robust to Adversarial Examples

    Shixiang Gu and Luca Rigazio. Towards Deep Neural Net- work Architectures Robust to Adversarial Examples. arXiv e-prints, page arXiv:1412.5068, Dec 2014

  4. [12]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceed- ings of the IEEE conference on computer vision and pattern recognition, pages 770–778, 2016

  5. [13]

    Squeeze-and-excitation net- works

    Jie Hu, Li Shen, and Gang Sun. Squeeze-and-excitation net- works. arXiv preprint arXiv:1709.01507, 7, 2017

  6. [14]

    Adam: A method for stochastic optimization

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 , 2014

  7. [15]

    Learning multiple layers of features from tiny images

    Alex Krizhevsky and Geoffrey Hinton. Learning multiple layers of features from tiny images. Technical report, Cite- seer, 2009

  8. [16]

    Imagenet classification with deep convolutional neural net- works

    Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convolutional neural net- works. In Advances in neural information processing sys- tems, pages 1097–1105, 2012

  9. [17]

    Ad- versarial examples in the physical world

    Alexey Kurakin, Ian Goodfellow, and Samy Bengio. Ad- versarial examples in the physical world. arXiv preprint arXiv:1607.02533, 2016

  10. [18]

    Ad- versarial Machine Learning at Scale

    Alexey Kurakin, Ian Goodfellow, and Samy Bengio. Ad- versarial Machine Learning at Scale. arXiv e-prints, page arXiv:1611.01236, Nov 2016

  11. [19]

    Defense against Adversarial At- tacks Using High-Level Representation Guided Denoiser

    Fangzhou Liao, Ming Liang, Yinpeng Dong, Tianyu Pang, Xiaolin Hu, and Jun Zhu. Defense against Adversarial At- tacks Using High-Level Representation Guided Denoiser. arXiv e-prints, page arXiv:1712.02976, Dec 2017

  12. [20]

    Feature pyramid networks for object detection

    Tsung-Yi Lin, Piotr Doll ´ar, Ross B Girshick, Kaiming He, Bharath Hariharan, and Serge J Belongie. Feature pyramid networks for object detection. In CVPR, volume 1, page 4, 2017

  13. [21]

    Deepfool: a simple and accurate method to fool deep neural networks

    Seyed-Mohsen Moosavi-Dezfooli, Alhussein Fawzi, and Pascal Frossard. Deepfool: a simple and accurate method to fool deep neural networks. In Proceedings of the IEEE Con- ference on Computer Vision and Pattern Recognition, pages 2574–2582, 2016

  14. [22]

    Rectified linear units im- prove restricted boltzmann machines

    Vinod Nair and Geoffrey E Hinton. Rectified linear units im- prove restricted boltzmann machines. In Proceedings of the 27th international conference on machine learning (ICML- 10), pages 807–814, 2010

  15. [23]

    Deep neural networks are easily fooled: High confidence predictions for unrecognizable images

    Anh Nguyen, Jason Yosinski, and Jeff Clune. Deep neural networks are easily fooled: High confidence predictions for unrecognizable images. In Proceedings of the IEEE con- ference on computer vision and pattern recognition , pages 427–436, 2015

  16. [24]

    Osadchy, J

    M. Osadchy, J. Hernandez-Castro, S. Gibson, O. Dunkel- man, and D. Prez-Cabo. No bot expects the deepcaptcha! introducing immutable adversarial examples, with applica- tions to captcha generation. IEEE Transactions on Informa- tion Forensics and Security, 12(11):2640–2653, Nov 2017

  17. [25]

    Berkay Celik, and Ananthram Swami

    Nicolas Papernot, Patrick McDaniel, Ian Goodfellow, Somesh Jha, Z. Berkay Celik, and Ananthram Swami. Prac- tical Black-Box Attacks against Machine Learning. arXiv e-prints, page arXiv:1602.02697, Feb 2016

  18. [26]

    Practi- cal black-box attacks against machine learning

    Nicolas Papernot, Patrick McDaniel, Ian Goodfellow, Somesh Jha, Z Berkay Celik, and Ananthram Swami. Practi- cal black-box attacks against machine learning. In Proceed- ings of the 2017 ACM on Asia Conference on Computer and Communications Security, pages 506–519. ACM, 2017

  19. [27]

    Towards the Science of Security and Privacy in Machine Learning

    Nicolas Papernot, Patrick McDaniel, Arunesh Sinha, and Michael Wellman. Towards the Science of Security and Privacy in Machine Learning. arXiv e-prints , page arXiv:1611.03814, Nov 2016

  20. [28]

    Generative adversarial perturbations

    Omid Poursaeed, Isay Katsman, Bicheng Gao, and Serge Be- longie. Generative adversarial perturbations. arXiv preprint arXiv:1712.02328, 2017

  21. [29]

    Faster r-cnn: Towards real-time object detection with region proposal networks

    Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster r-cnn: Towards real-time object detection with region proposal networks. In Advances in neural information pro- cessing systems, pages 91–99, 2015

  22. [30]

    Very deep convo- lutional networks for large-scale image recognition

    Karen Simonyan and Andrew Zisserman. Very deep convo- lutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556, 2014

  23. [31]

    One pixel attack for fooling deep neural networks

    Jiawei Su, Danilo Vasconcellos Vargas, and Kouichi Sakurai. One pixel attack for fooling deep neural networks. IEEE Transactions on Evolutionary Computation, 2019

  24. [32]

    Intriguing properties of neural networks

    Christian Szegedy, Wojciech Zaremba, Ilya Sutskever, Joan Bruna, Dumitru Erhan, Ian Goodfellow, and Rob Fergus. Intriguing properties of neural networks. arXiv preprint arXiv:1312.6199, 2013

  25. [33]

    Deepface: Closing the gap to human-level perfor- mance in face verification

    Yaniv Taigman, Ming Yang, Marc’Aurelio Ranzato, and Lior Wolf. Deepface: Closing the gap to human-level perfor- mance in face verification. In Proceedings of the IEEE con- ference on computer vision and pattern recognition , pages 1701–1708, 2014

  26. [34]

    Ensemble Adversarial Training: Attacks and Defenses

    Florian Tram `er, Alexey Kurakin, Nicolas Papernot, Ian Goodfellow, Dan Boneh, and Patrick McDaniel. Ensemble Adversarial Training: Attacks and Defenses. arXiv e-prints, page arXiv:1705.07204, May 2017

  27. [35]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszko- reit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. In Advances in Neural Information Processing Systems, pages 5998–6008, 2017

  28. [36]

    Generating adversarial examples with adversarial networks

    Chaowei Xiao, Bo Li, Jun-Yan Zhu, Warren He, Mingyan Liu, and Dawn Song. Generating adversarial examples with adversarial networks. arXiv preprint arXiv:1801.02610, 2018

  29. [2009]

    IEEE Conference on , pages 248–255

    CVPR 2009. IEEE Conference on , pages 248–255. Ieee, 2009

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.