- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 1
- Soundness
- 2
Read review
Paper summary
This paper proposes a class-wise adversarial vulnerability assessment framework based solely on clean data. By analyzing the softmax prediction probabilities of a model on a clean dataset, the framework aims to estimate the adversarial vulnerability of different classes. Specifically, the authors introduce two metrics, SOIV-U and SOIV-T. SOIV-U uses the ideal-centered second-order moment of the true-class prediction probability to assess class-wise vulnerability under untargeted attacks, while SOIV-T employs the cross second order moment of source class and target class prediction probabilities to characterize directional vulnerability under targeted attacks. Experimentally, the authors evaluate the proposed framework across multiple datasets using various adversarial attack methods and report correlations between the proposed metrics and the observed class-wise ASRs.
Critical strengths
The SOIV framework proposed in this paper only relies on the Softmax prediction probabilities on a clean validation set to obtain class-level and source-to-target directional vulnerability statistics. The authors report about a 162x speed-up in computation under specific experimental settings. This low-cost evaluation method is suitable for model pre-screening scenarios where you need to quickly identify potentially vulnerable classes.
Critical weaknesses
-
Some relevant literature is missing. Existing studies have separately explored robustness evaluation without attacks [1], class level differences in adversarial robustness and adaptive defenses [2], as well as source-to-target class attack tendencies [3]. Although this paper proposes specific second order probabilistic statistical metrics, their incremental contribution compared to these existing studies has not been fully clarified. In addition, SOIV-T lacks a direct comparison with simple source-to-target conditional probability statistics, so the independent value of its directional vulnerability estimation is still unclear.
-
This paper mainly verifies the effectiveness of the SOIV and ASR correlation methods, but in the targeted attack experiments, only one source class is chosen for each dataset and model, making it hard to prove that the proposed metrics can accurately capture the full vulnerability relationships between classes. Also, the authors choose the attack intensity based on the class-level ASR distribution, and Appendix D.7 shows that the correlation changes noticeably with attack intensity. So, the current experiments aren’t enough to show that SOIV can provide a stable and reliable vulnerability assessment under different attack conditions.
[1] Attack-free evaluating and enhancing adversarial robustness on categorical data. [2] To be Robust or to be Fair: Towards Fairness in Adversarial Training. [3] Improving robust fairness via balance adversarial training.
Questions for authors
-
In Section 3, the author estimates adversarial vulnerability through statistics of predicted probabilities on clean samples. However, SOIV-U includes both prediction variance and confidence bias, while SOIV-T is closely related to category confusion of normal samples. Could the author further explain why these metrics can reflect adversarial vulnerability, rather than just normal classification difficulty, and demonstrate their added value compared to simple confidence statistics?
-
In Section 4.2, the author only validates SOIV-T using Automobile as the source category, yet in Section 4.3, they present a complete source-target vulnerability matrix. Could the author explain why this experiment is enough to support conclusions about the overall directional vulnerability structure, and provide validation results for more source categories?
-
In Section 4.1, the author verifies the effectiveness of SOIV through its correlation with various attack ASRs. However, different attacks’ optimization strengths and perturbation constraints might affect ASR and class ranking. Could the author clarify whether the attack settings used were fully optimized, and verify if SOIV can still maintain stable correlation under a unified threat model and more fully optimized attack conditions?
5 comments
Response to Question 2 (Source-Class Validation)
We agree that evaluating only Automobile was insufficient to establish generality. We have therefore extended the targeted evaluation to the remaining nine CIFAR-10 source classes. The table reports the correlation between and targeted ASR across the nine possible target classes for each source. Each entry represents Pearson/Spearman/Kendall.
| Source | FGSM | PGD | CW | MIXED |
|---|---|---|---|---|
| Airplane | 0.701/0.783/0.556 | 0.675/0.817/0.611 | 0.719/0.817/0.611 | 0.726/0.850/0.667 |
| Bird | 0.647/0.661/0.535 | 0.757/0.667/0.556 | 0.762/0.633/0.500 | 0.743/0.650/0.556 |
| Cat | 0.575/0.767/0.667 | 0.650/0.800/0.722 | 0.771/0.733/0.667 | 0.675/0.800/0.722 |
| Deer | 0.549/0.717/0.556 | 0.707/0.750/0.556 | 0.679/0.717/0.556 | 0.654/0.700/0.500 |
| Dog | 0.801/0.895/0.761 | 0.697/0.900/0.778 | 0.832/0.850/0.722 | 0.788/0.900/0.778 |
| Frog | 0.804/0.733/0.556 | 0.739/0.800/0.667 | 0.796/0.700/0.500 | 0.803/0.733/0.556 |
| Horse | 0.823/0.900/0.778 | 0.829/0.850/0.722 | 0.821/0.900/0.778 | 0.826/0.900/0.778 |
| Ship | 0.935/0.613/0.514 | 0.779/0.733/0.611 | 0.889/0.617/0.556 | 0.881/0.733/0.667 |
| Truck | 0.902/0.750/0.556 | 0.918/0.800/0.667 | 0.958/0.833/0.667 | 0.937/0.817/0.722 |
All correlations are positive across the additional source classes. Under MIXED attacks, Pearson ranges from 0.654 to 0.937 and Spearman from 0.650 to 0.900. These results support directional alignment beyond Automobile, although its strength varies by source class.
Response to Question 3 (Attack Settings)
We thank the reviewer for raising concerns about attack strength and optimization. The manuscript already provides evidence across attacks in Section 4.1 (Table 1), where SOIV-U is compared with FGSM, PGD, C&W, Square, AutoAttack, and MIXED. In particular, standard AutoAttack (, ) yields Pearson, Spearman, and Kendall correlations of 0.933, 0.915, and 0.778, respectively. Also, MIXED yields 0.927, 0.915, and 0.778. Appendix B.5 (Table 6) specifies the attack configurations, and all ASRs are measured on clean-correct test samples. Thus, the original evaluation already includes a standardized AutoAttack assessment rather than relying only on individually configured attacks.
The manuscript also examines why attack strength affects correlation. Section 3.4 and Appendix A.1 motivate SOIV through a conditional connection between clean probability variability and local adversarial sensitivity, based on a first-order approximation that may become less accurate for larger perturbations. Appendix D.7 (Table 39) explicitly examines stronger attacks and the spread of class-wise ASRs. As the PGD perturbation budget increases from to , , and , the standard deviation of class-wise ASR decreases from 14.46 to 5.95, 1.01, and 0 percentage points, respectively. At the strongest PGD and C&W settings, all classes reach 100% ASR, making the correlation undefined because no meaningful class-wise ranking remains. This does not necessarily indicate instability in SOIV itself. For a fixed model and clean validation set, SOIV values remain unchanged, whereas the attack-based reference becomes saturated and loses its ability to distinguish vulnerable classes. However, we acknowledge that saturation does not explain every decline in correlation, since stronger attacks may also change relative class rankings. As explained in Appendix B.5, attack budgets were therefore selected to preserve measurable class-wise ASR differences, allowing meaningful evaluation of relative vulnerability.
To test whether the observed alignment depends on the original hyperparameters, we additionally evaluated CIFAR-10/ResNet-18 under a unified budget of and varied PGD optimization steps and perturbation budgets
| Attack (, steps) | Pearson | Spearman | Kendall |
|---|---|---|---|
| FGSM (1.5/255) | 0.959 | 0.939 | 0.822 |
| Square (1.5/255) | 0.963 | 0.952 | 0.867 |
| AutoAttack (1.5/255) | 0.933 | 0.915 | 0.778 |
| PGD (1.5/255, 20) | 0.935 | 0.915 | 0.778 |
| PGD (1.5/255, 50) | 0.937 | 0.915 | 0.778 |
| PGD (1.5/255, 100) | 0.937 | 0.915 | 0.778 |
| PGD (2.5/255, 10) | 0.824 | 0.855 | 0.733 |
| PGD (4/255, 10) | 0.728 | 0.806 | 0.689 |
The original six-step PGD setting (, Table 1) achieved Pearson, Spearman, and Kendall correlations of 0.936, 0.939, and 0.822, respectively. Increasing PGD to 20–100 steps under the same perturbation budget produced broadly similar correlations, while larger budgets reduced the correlations without reversing their signs in the additional experiments. Nevertheless, increasing the number of PGD steps does not guarantee global attack optimality. Furthermore, Appendix D.8 (Table 40) evaluates nine combinations of data split and training seeds, reporting mean SOIV-U correlations of 0.870, 0.888, and 0.738 for Pearson, Spearman, and Kendall against MIXED ASR, respectively. These results provide additional evidence of reproducibility across random seeds, although they do not establish invariance across threat models. Overall, SOIV-U remains useful for relative class-wise vulnerability assessment when meaningful ASR differences exist, rather than guaranteeing consistent rankings under arbitrary attack strengths or saturated attack conditions.
Response to Weakness 1 & Question 1 (Novelty)
We thank the reviewer for highlighting [1–3], which study related but distinct aspects of adversarial robustness. Specifically, [1] introduces gradient-based IGSG to evaluate feature sensitivity on categorical data. [2] examines class-wise robustness disparities arising from adversarial training and proposes a fairness-oriented method. [3] analyzes source-class attack difficulty and target-class bias to improve adversarial training. In contrast, our work estimates class-wise and source-to-target vulnerability in image classifiers using only clean softmax outputs, without generating adversarial examples or computing input gradients. Our contribution is not the discovery of class imbalance or attack directionality, but an attack-free approach to characterizing their relative vulnerability structure before performing attacks.
The adversarial relevance of SOIV-U follows from its ideal-centered formulation
E[(1−ps)2∣Y=s]=Var(ps∣Y=s)+(1−E[ps∣Y=s])2.
Thus, SOIV-U captures both within-class confidence variability and the squared mean confidence deficit. Section 3.4 and Appendix A.1 provide a conditional theoretical motivation for its relationship with adversarial vulnerability. Under local smoothness and sufficient coverage of probability-sensitive directions by natural within-class variations, clean probability variability and small-perturbation sensitivity depend on the same local probability gradient. This supports SOIV-U as a relative vulnerability indicator without implying equivalence to ASR or a robustness guarantee.
Ideal centering also addresses a limitation of conventional sample variance. Since variance is centered at each class's empirical mean, it can become small even when true-class probabilities are consistently close to zero. In contrast, SOIV-U measures deviations from the ideal confidence of one. Appendix D.4 demonstrates this distinction experimentally on adversarially trained CIFAR-100 models. Sample variance exhibits negative correlations with ASR across untargeted attacks (Tables 32 and 34), whereas SOIV-U maintains positive correlations (Tables 19 and 21). This provides a concrete advantage in assessing adversarial vulnerability beyond simply using a different statistical formulation.
We further compare SOIV-U with simpler statistics on CIFAR-10 using ResNet-18. All statistics are computed from 5,000 clean validation images, while ASR is evaluated on a separate test set of 10,000 images. The MIXED untargeted setting aggregates FGSM, PGD, C&W, Square, and AutoAttack (Section 4.1, Table 1). Here, Variance denotes the unbiased within-class variance of true-class probabilities, while Confidence is defined as 1−E[ps∣Y=s] so that larger values indicate greater vulnerability.
Parentheses indicate two-sided p-values. SOIV-U matches or exceeds both baselines here, without implying universal predictive superiority.
For directional vulnerability, the magnitude of SOIV-T is
E[(1−ps)pt∣Y=s].
SOIV-T measures directional vulnerability by jointly considering the source-confidence deficit and target probability, ∣SOIV-T(s,t)∣=E[(1−ps)pt∣Y=s], whereas conditional target probability, E[pt∣Y=s], uses only the latter. We compare both metrics on CIFAR-10/ResNet-18 using airplane as the source class and MIXED targeted ASR from FGSM, PGD, and C&W.
SOIV-T achieves higher correlations across all three measures, particularly for Spearman and Kendall. These results suggest that incorporating the source-confidence deficit can improve alignment with targeted ASR in this setting. The second-order formulation also supports the unified variance–covariance analysis in Section 4.3.
Finally, Section 5.2 (Table 3) demonstrates practical relevance to defense. SOIV-U-guided adversarial training allocates adversarial examples by estimated class-wise vulnerability rather than uniformly. Under the same budget, overall AutoAttack ASR falls from 61.43% to 58.17% (L∞, ϵ=8/255), improving 7 of 10 classes. This single setting does not establish superiority over [2,3], but shows how SOIV estimates can guide defense resource allocation. Overall, SOIV offers an efficient, attack-free structural diagnostic rather than universally superior ASR prediction.