Full search page →

Athalye et al., “Obfuscated Gradients Give a False Sense of Security” (ICML 2018)

Optional · paper guide 2 of 3 for DLCV 14 · Robustness: Distribution Shift, Adversarial Examples and Anomalies. Citation numbers refer to the reference list at the bottom of this page.

Navigate: this page and this track: Navigate: this page and this track

On this page

Supplementary notes · 6

  1. Effective robustness and accuracy on the line
  2. Test-time entropy minimization
  3. Domain-adversarial training and the gradient-reversal layer
  4. Attack settings that decide an evaluation
  5. Randomized smoothing in practice
  6. PatchCore step by step

Paper guides · 3

  1. Madry et al., “Towards Deep Learning Models Resistant to Adversarial Attacks” (ICLR 2018)
  2. Athalye et al., “Obfuscated Gradients Give a False Sense of Security” (ICML 2018)
  3. Gulrajani and Lopez-Paz, “In Search of Lost Domain Generalization” (ICLR 2021)

Deep Learning for Computer Vision

  1. 02 · Deep Image Classification
View the whole course tree →

Challenge

Many defenses reported high robustness to white-box iterative attacks. Were they robust, or did they only prevent the attacker’s gradient from being useful [29]?

Teaching evaluation diagram — attack failure triggers a diagnosis and an adaptive attack; success or failure is checked on the actual defense under its original threat model.
Teaching evaluation diagram — attack failure triggers a diagnosis and an adaptive attack; success or failure is checked on the actual defense under its original threat model.

Method

Identify the mechanism, gradient masking in the form of obfuscated gradients, and build attacks that do not rely on the masked gradient. The paper names three types (shattered, stochastic, and vanishing or exploding gradients), lists behaviors that reveal them (the warning signs in the main text), and for each type gives a counter-attack: BPDA, EOT and reparameterization.

Results

In a case study of the non-certified white-box defenses accepted at ICLR 2018, seven of nine relied on obfuscated gradients; the new attacks circumvented six completely and one partially, in the threat model each paper had considered [29]. Two certified defenses and one with only black-box claims were left out of the case study. The paper also stresses an asymmetry: to break a defense one attack suffices, while a defender must show that no attack within the threat model succeeds.

Reading the evidence

Read Section 3 (the three types and their symptoms) and Section 4 (the attacks) first, then Table 1, which lists each defense with its dataset, its perturbation bound and the accuracy left under the new attacks. Note the two adversarial-training defenses at the bottom of the table: the authors found that Madry et al.’s PGD training does not rely on obfuscated gradients. Check: (a) that each attack keeps the original paper’s threat model and ε\varepsilon, so the numbers are a fair re-evaluation; (b) the per-defense sections, which show how BPDA’s approximation was chosen in each case; (c) the evaluation recommendations in the discussion.

Problems

The analysis is a case study of one conference’s defenses; it shows that these defenses fail, not that all defenses of a type must fail. Adaptive attacks require effort and expertise, which is the gap that AutoAttack later tried to close for standard settings [30].

Conclusion

It is the reason this chapter asks for white-box, adaptive and ensemble attacks before any robustness claim is believed, and the source of the warning signs used in Exercise 2.

Following an adaptive gradient

BPDA keeps the defense’s actual forward transformation h(x)h(x) but substitutes a differentiable approximation gg in the backward pass. If vv is the loss gradient arriving at that transformation’s output, the approximate input gradient is Jg(x)⊤vJ_g(x)^\top v, where JgJ_g is the Jacobian of gg. For example, a rounding operation can remain in the forward pass while an identity approximation supplies its backward derivative. The attacker must evaluate the resulting candidate on the real rounded model, not on the surrogate alone [29].

For a randomized transformation τ\tau, EOT estimates the expected-loss gradient with samples:

∇xEτL(f(τ(x)),y)≈1K∑k=1K∇xL(f(τk(x)),y).\nabla_x\mathbb E_{\tau}\mathcal L(f(\tau(x)),y) \approx\frac1K\sum_{k=1}^{K}\nabla_x\mathcal L(f(\tau_k(x)),y).

ff is the defended predictor, yy the target label used by the attack objective L\mathcal L, KK the sample count and τk\tau_k sampled transformations. Differentiation/exchange of expectation needs the relevant regularity; nondifferentiable transformations may additionally require a backward approximation. Increasing attack effort changes the reliability of this empirical probe, not the permitted perturbation set [29].

References

  1. D. Hendrycks, K. Zhao, S. Basart, J. Steinhardt and D. Song, “Natural Adversarial Examples,” CVPR, 2021. CVF
  2. P. W. Koh, S. Sagawa, H. Marklund, S. M. Xie, M. Zhang, A. Balsubramani et al., “WILDS: A Benchmark of in-the-Wild Distribution Shifts,” ICML, PMLR 139, 2021. PMLR
  3. J. Yang, K. Zhou, Y. Li and Z. Liu, “Generalized Out-of-Distribution Detection: A Survey,” International Journal of Computer Vision, vol. 132, no. 12, pp. 5635–5662, 2024. doi · arXiv
  4. D. Hendrycks and T. Dietterich, “Benchmarking Neural Network Robustness to Common Corruptions and Perturbations,” ICLR, 2019. arXiv
  5. D. Hendrycks, S. Basart, N. Mu, S. Kadavath, F. Wang, E. Dorundo et al., “The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization,” ICCV, 2021. CVF
  6. A. Barbu, D. Mayo, J. Alverio, W. Luo, C. Wang, D. Gutfreund, J. Tenenbaum and B. Katz, “ObjectNet: A large-scale bias-controlled dataset for pushing the limits of object recognition models,” NeurIPS, 2019. proceedings
  7. B. Recht, R. Roelofs, L. Schmidt and V. Shankar, “Do ImageNet Classifiers Generalize to ImageNet?,” ICML, PMLR 97, 2019. PMLR
  8. R. Taori, A. Dave, V. Shankar, N. Carlini, B. Recht and L. Schmidt, “Measuring Robustness to Natural Distribution Shifts in Image Classification,” NeurIPS, 2020. proceedings
  9. J. Miller, R. Taori, A. Raghunathan, S. Sagawa, P. W. Koh, V. Shankar, P. Liang, Y. Carmon and L. Schmidt, “Accuracy on the Line: on the Strong Correlation Between Out-of-Distribution and In-Distribution Generalization,” ICML, PMLR 139, 2021. PMLR
  10. D. Hendrycks, N. Mu, E. D. Cubuk, B. Zoph, J. Gilmer and B. Lakshminarayanan, “AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty,” ICLR, 2020. arXiv
  11. R. Geirhos, P. Rubisch, C. Michaelis, M. Bethge, F. A. Wichmann and W. Brendel, “ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness,” ICLR, 2019. arXiv
  12. A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal et al., “Learning Transferable Visual Models From Natural Language Supervision,” ICML, PMLR 139, 2021. PMLR
  13. A. Fang, G. Ilharco, M. Wortsman, Y. Wan, V. Shankar, A. Dave and L. Schmidt, “Data Determines Distributional Robustness in Contrastive Language Image Pre-training (CLIP),” ICML, PMLR 162, 2022. PMLR
  14. M. Wortsman, G. Ilharco, J. W. Kim, M. Li, S. Kornblith, R. Roelofs et al., “Robust Fine-Tuning of Zero-Shot Models,” CVPR, 2022. CVF
  15. S. Schneider, E. Rusak, L. Eck, O. Bringmann, W. Brendel and M. Bethge, “Improving robustness against common corruptions by covariate shift adaptation,” NeurIPS, 2020. proceedings
  16. D. Wang, E. Shelhamer, S. Liu, B. Olshausen and T. Darrell, “Tent: Fully Test-Time Adaptation by Entropy Minimization,” ICLR, 2021. arXiv
  17. Y. Ganin, E. Ustinova, H. Ajakan, P. Germain, H. Larochelle, F. Laviolette, M. Marchand and V. Lempitsky, “Domain-Adversarial Training of Neural Networks,” Journal of Machine Learning Research, vol. 17, no. 59, pp. 1–35, 2016. JMLR
  18. K. Zhou, Z. Liu, Y. Qiao, T. Xiang and C. C. Loy, “Domain Generalization: A Survey,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022. doi · arXiv
  19. I. Gulrajani and D. Lopez-Paz, “In Search of Lost Domain Generalization,” ICLR, 2021. arXiv
  20. Y. Ovadia, E. Fertig, J. Ren, Z. Nado, D. Sculley, S. Nowozin, J. V. Dillon, B. Lakshminarayanan and J. Snoek, “Can you trust your model’s uncertainty? Evaluating predictive uncertainty under dataset shift,” NeurIPS, 2019. proceedings
  21. C. Szegedy, W. Zaremba, I. Sutskever, J. Bruna, D. Erhan, I. Goodfellow and R. Fergus, “Intriguing properties of neural networks,” ICLR, 2014. arXiv
  22. I. J. Goodfellow, J. Shlens and C. Szegedy, “Explaining and Harnessing Adversarial Examples,” ICLR, 2015. arXiv
  23. A. Madry, A. Makelov, L. Schmidt, D. Tsipras and A. Vladu, “Towards Deep Learning Models Resistant to Adversarial Attacks,” ICLR, 2018. arXiv
  24. F. Croce, M. Andriushchenko, V. Sehwag, E. Debenedetti, N. Flammarion, M. Chiang, P. Mittal and M. Hein, “RobustBench: a standardized adversarial robustness benchmark,” NeurIPS Datasets and Benchmarks Track, 2021. proceedings · leaderboard
  25. N. Papernot, P. McDaniel, I. Goodfellow, S. Jha, Z. B. Celik and A. Swami, “Practical Black-Box Attacks against Machine Learning,” ACM Asia Conference on Computer and Communications Security (AsiaCCS), pp. 506–519, 2017. doi
  26. N. Carlini and D. Wagner, “Towards Evaluating the Robustness of Neural Networks,” IEEE Symposium on Security and Privacy (SP), pp. 39–57, 2017. doi
  27. D. Tsipras, S. Santurkar, L. Engstrom, A. Turner and A. Madry, “Robustness May Be at Odds with Accuracy,” ICLR, 2019. arXiv
  28. H. Zhang, Y. Yu, J. Jiao, E. P. Xing, L. El Ghaoui and M. I. Jordan, “Theoretically Principled Trade-off between Robustness and Accuracy,” ICML, PMLR 97, 2019. PMLR
  29. A. Athalye, N. Carlini and D. Wagner, “Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examples,” ICML, PMLR 80, 2018. PMLR
  30. F. Croce and M. Hein, “Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attacks,” ICML, PMLR 119, 2020. PMLR
  31. J. Cohen, E. Rosenfeld and J. Z. Kolter, “Certified Adversarial Robustness via Randomized Smoothing,” ICML, PMLR 97, 2019. PMLR
  32. K. Eykholt, I. Evtimov, E. Fernandes, B. Li, A. Rahmati, C. Xiao, A. Prakash, T. Kohno and D. Song, “Robust Physical-World Attacks on Deep Learning Visual Classification,” CVPR, 2018. CVF
  33. T. B. Brown, D. Mané, A. Roy, M. Abadi and J. Gilmer, “Adversarial Patch,” arXiv:1712.09665, 2017. arXiv
  34. D. Hendrycks and K. Gimpel, “A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks,” ICLR, 2017. arXiv
  35. W. Liu, X. Wang, J. D. Owens and Y. Li, “Energy-based Out-of-distribution Detection,” NeurIPS, 2020. proceedings
  36. K. Lee, K. Lee, H. Lee and J. Shin, “A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attacks,” NeurIPS, 2018. proceedings
  37. Y. Sun, Y. Ming, X. Zhu and Y. Li, “Out-of-Distribution Detection with Deep Nearest Neighbors,” ICML, PMLR 162, 2022. PMLR
  38. D. Hendrycks, M. Mazeika and T. Dietterich, “Deep Anomaly Detection with Outlier Exposure,” ICLR, 2019. arXiv
  39. J. Zhang, J. Yang, P. Wang, H. Wang, Y. Lin, H. Zhang et al., “OpenOOD v1.5: Enhanced Benchmark for Out-of-Distribution Detection,” arXiv:2306.09301, 2023. arXiv
  40. P. Bergmann, M. Fauser, D. Sattlegger and C. Steger, “MVTec AD — A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection,” CVPR, 2019. CVF
  41. L. Heckler-Kram, J.-H. Neudeck, U. Scheler, R. König and C. Steger, “The MVTec AD 2 Dataset: Advanced Scenarios for Unsupervised Anomaly Detection,” International Journal of Computer Vision, vol. 134, no. 4, 2026. doi · arXiv
  42. K. Roth, L. Pemula, J. Zepeda, B. Schölkopf, T. Brox and P. Gehler, “Towards Total Recall in Industrial Anomaly Detection,” CVPR, 2022. CVF
  43. J. Bitterwolf, M. Müller and M. Hein, “In or Out? Fixing ImageNet Out-of-Distribution Detection Evaluation,” ICML, PMLR 202, 2023. PMLR
  44. J. Liang, R. He and T. Tan, “A Comprehensive Survey on Test-Time Adaptation Under Distribution Shifts,” International Journal of Computer Vision, vol. 133, no. 1, pp. 31–64, 2025. doi · arXiv
  45. N. Akhtar, A. Mian, N. Kardan and M. Shah, “Advances in Adversarial Attacks and Defenses in Computer Vision: A Survey,” IEEE Access, vol. 9, pp. 155161–155196, 2021. doi · arXiv
  46. G. Pang, C. Shen, L. Cao and A. van den Hengel, “Deep Learning for Anomaly Detection: A Review,” ACM Computing Surveys, vol. 54, no. 2, pp. 1–38, 2021. doi · arXiv
  47. J. Liu, G. Xie, J. Wang, S. Li, C. Wang, F. Zheng and Y. Jin, “Deep Industrial Image Anomaly Detection: A Survey,” Machine Intelligence Research, vol. 21, no. 1, pp. 104–135, 2024. doi · arXiv
  48. Z. Wang, T. Pang, C. Du, M. Lin, W. Liu and S. Yan, “Better Diffusion Models Further Improve Adversarial Training,” ICML, PMLR 202, 2023. PMLR
  49. J. Yang, P. Wang, D. Zou, Z. Zhou, K. Ding, W. Peng et al., “OpenOOD: Benchmarking Generalized Out-of-Distribution Detection,” NeurIPS Datasets and Benchmarks Track, 2022. proceedings
  50. J. Jeong, Y. Zou, T. Kim, D. Zhang, A. Ravichandran and O. Dabeer, “WinCLIP: Zero-/Few-Shot Anomaly Classification and Segmentation,” CVPR, 2023. CVF