- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 1
- Soundness
- 2
Read review
Paper summary
This paper proposes C2FDA, a diffusion-based framework for local black-box adversarial attacks on time series classifiers. The method employs a coarse-to-fine attention mechanism to jointly determine perturbation locations and generate perturbation content, together with a hybrid boundary objective to improve attack effectiveness. Experiments on demonstrate promising attack success rates with limited temporal coverage and low perturbation distortion.
Critical strengths
- The paper addresses the problem by jointly considering perturbation locations and content in local time series attacks.
- The proposed coarse to fine attention mechanism provides a coherent framework for coordinating temporal localization and perturbation generation.
- Extensive experiments across 14 datasets and five target models demonstrate strong attack performance with limited temporal coverage.
Critical weaknesses
- The generalizability of the attack remains uncertain. The evaluation relies on a single surrogate model and relatively lightweight target classifiers, most of which are based on conventional architectures. The lack of experiments on modern time series classification models, such as Transformer-based models, limits the evidence for the broader applicability of the proposed method.
- Insufficient validation of the joint where–what learning. A central claim of the paper is that jointly optimizing perturbation locations and content is more effective than treating them separately. However, the ablation studies mainly evaluate individual attention components and do not directly compare joint optimization with a decoupled alternative using the same backbone. Such a comparison would better substantiate the claimed advantage of the proposed coupling mechanism.
- The overall presentation needs further refinement. The writing would benefit from additional language polishing. Figures 1 and 2 could be improved in terms of visual clarity and consistency, while Tables 2, 3, and 4 are overly dense and could benefit from better formatting and readability.
Questions for authors
See Weaknesses
9 comments
W1
W1 (continued).
Surrogates. With CNN as the target, Table 6 compares C2FDA and SGAN over 14 datasets. C2FDA's average ASR / MSE is 73.79% / 0.0379 with MLP, 82.72% / 0.0334 with ResNet, and 86.85% / 0.0322 with a Transformer, versus SGAN's 46.2% / 0.0749, 53.7% / 0.0686, and 59.2% / 0.0757, respectively; C2FDA reaches 86.85% average ASR with the Transformer surrogate, compared with 82.72% with the default ResNet surrogate. (Details: Sec. 4.5, Table 6.)
| Dataset | MLP C2FDA (ASR (%) / MSE) | MLP SGAN (ASR (%) / MSE) | ResNet C2FDA (ASR (%) / MSE) | ResNet SGAN (ASR (%) / MSE) | Transformer C2FDA (ASR (%) / MSE) | Transformer SGAN (ASR (%) / MSE) |
|---|---|---|---|---|---|---|
| BasicMotions | / 0.0724 | 50.0 / | / 0.0606 | 56.2 / | 68.8 / 0.0571 | / |
| HandMovementDirection | 77.8 / 0.0411 | / | 94.4 / 0.0342 | / | 88.9 / 0.0338 | / |
| UWaveGestureLibrary | 89.6 / | / 0.0926 | 97.4 / 0.0845 | / | 93.5 / 0.0813 | / |
| Car | 80.0 / 0.0148 | / | 86.7 / 0.0126 | / | 93.3 / 0.0124 | / |
| CinCECGTorso | / 0.0121 | 70.1 / | / 0.0104 | 79.9 / | / 0.0098 | 84.2 / |
| ECG5000 | 89.2 / 0.0243 | / | 100.0 / 0.0221 | / | 98.7 / 0.0219 | / |
| ECGFiveDays | 93.2 / 0.0207 | / | 100.0 / | / 0.0172 | 98.3 / 0.0181 | / |
| Earthquakes | 83.1 / 0.1031 | 0.0 / — | 94.8 / 0.0944 | 0.0 / — | 96.1 / 0.0896 | 0.0 / — |
| FordA | 79.3 / 0.0156 | / | 87.3 / 0.0127 | / | 92.5 / 0.0118 | / |
| FordB | 82.6 / 0.0142 | / | 90.3 / 0.0127 | / | 89.0 / 0.0139 | / |
| Trace | 62.5 / 0.0109 | / | 72.5 / 0.0092 | / | 82.5 / 0.0086 | / |
| ItalyPowerDemand | 78.4 / 0.0445 | / | 87.3 / 0.0403 | / | 91.1 / 0.0386 | / |
| SonyAIBORobotSurface1 | 62.9 / 0.0287 | / | 71.8 / 0.0250 | / | 78.2 / 0.0237 | / |
| SonyAIBORobotSurface2 | 46.7 / 0.0348 | / | 53.3 / 0.0319 | / | 65.1 / 0.0307 | / |
| Average (14 datasets) | 73.79 / 0.0379 | / | 82.72 / 0.0334 | / | 86.85 / 0.0322 | / |
W2 & W3
W2. "Insufficient validation of the joint where–what learning."
We added two decoupled alternatives with the same backbone and C2F modules, and the full C2FDA beats both on every dataset and metric: Simple Joint still updates the WHERE and WHAT heads simultaneously with the same Top-k mask and hybrid loss but removes the WHERE→WHAT conditioning, and Sequential keeps the full architecture but learns WHERE first, fixes the mask, and then learns WHAT.
Joint versus decoupled. Table 3 reports the CNN-target ablations:
| Dataset | Metric | Full C2FDA | w/o Coarse | w/o Fine | w/o C2F Attention | Random Intervals | Exact Margin Only | Simple Joint | Sequential |
|---|---|---|---|---|---|---|---|---|---|
| BasicMotions | ASR(%)↑ | 31.2 | 56.2 | 31.2 | 43.8 | 43.8 | 37.5 | ||
| MSE↓ | 0.0897 | 0.0504 | 0.1915 | 0.0979 | 0.0872 | 0.0814 | 0.0768 | ||
| DTW↓ | 71.74 | 97.83 | 239.27 | 102.61 | 94.38 | 96.17 | 89.21 | ||
| HandMovementDirection | ASR(%)↑ | 94.4 | 77.8 | 66.7 | 50.0 | 83.3 | 72.2 | 61.1 | |
| MSE↓ | 0.0608 | 0.0302 | 0.2046 | 0.0573 | 0.0415 | 0.0427 | 0.0684 | ||
| DTW↓ | 103.72 | 90.34 | 504.85 | 143.72 | 97.63 | 103.57 | 116.92 | ||
| UWaveGestureLibrary | ASR(%)↑ | 97.4 | 76.6 | 84.4 | 49.4 | 87.0 | 84.4 | 80.5 | |
| MSE↓ | 0.1024 | 0.0842 | 0.1490 | 0.0962 | 0.1183 | 0.0967 | 0.0971 | ||
| DTW↓ | 48.21 | 57.67 | 248.41 | 56.38 | 61.85 | 60.42 | 63.84 |
Full C2FDA reaches 50.0%, 94.4%, and 97.4% ASR, versus 43.8%, 72.2%, and 84.4% for Simple Joint and 37.5%, 61.1%, and 80.5% for Sequential, with lower MSE and DTW in every case. Since each control changes one factor, both the coupling and joint optimization contribute; even without coupling, simultaneous optimization (Simple Joint) exceeds Sequential in ASR on all three datasets. (Details: Sec. 4.3 "Effect of joint where–what learning", Table 3.)
Attention ablations. As the reviewer notes, w/o Coarse, w/o Fine, and w/o C2F Attention evaluate components. Removing C2F attention also changes feature extraction and conditioning paths, so we report it as a module ablation; Simple Joint and Sequential provide the direct coupling tests while retaining the same backbone and C2F modules. (Details: Sec. 4.3; App. D, Table 11 (9).)
W3. "The overall presentation needs further refinement."
We polished the language throughout, revised Figures 1 and 2, and reformatted the dense tables, while preserving the reported numerical values in unchanged comparisons.
Language. We proofread the main text and appendix, fixing grammar and improving sentence structure, making comparisons explicit, and unifying terminology.
Figures 1 and 2. Each perturbed segment now starts and ends on the original series, so the perturbed series is continuous. In Figure 1, the original and perturbed series boxes share one size and position across panels, panel (b) uses the same split arrows as panel (a), and titles and label colors are consistent. Figure 2 follows the same convention (black clean series, red perturbed support joined to it); its binary mask marks the same two windows, the Product operator is a readable block like Top-k, "Exact budget" matches the selection in Sec. 3.3, and a one-off green arrow now uses the WHAT color. (Details: Fig. 1, p. 2; Fig. 2, p. 4.)
Tables. Table 2 keeps all attacks, targets, and metrics in one table for direct comparison and is now easier to read: ASR cells drop the repeated percent sign, dataset names are rotated, target groups are separated, and C2FDA rows are shaded, which together allow a larger font at the same height. Table 3 uses a fixed 9-pt font, Tables 4 and 6 give each metric its own column instead of slash-separated cells, and Table 5 uses full dataset names. (Details: Tables 2–6, pp. 7–10.)
We hope these answers address the reviewer's concerns and would be glad to clarify any remaining point.
W1
We thank the reviewer for the constructive review. Each point below opens with a one-sentence answer and then follows the reviewer's sub-points in order. Table and figure numbers refer to the revised PDF, with submitted numbers in parentheses where they differ. In the revised PDF, revisions for this review are in orange and revisions shared with other reviews are in purple.
W1. "The evaluation relies on a single surrogate model and relatively lightweight target classifiers, most of which are based on conventional architectures."
Against a held-out Transformer target [1], C2FDA has the highest ASR of 13 attacks on all three multivariate datasets. Across three surrogate architectures with CNN as target, C2FDA also has higher 14-dataset average ASR and lower average MSE than SGAN for each surrogate.
Transformer target. For the three multivariate datasets, we evaluate transfer to a held-out Transformer target [1]. C2FDA reaches 62.5%, 66.7%, and 72.5% ASR, versus 37.5% (SGAN), 50.0% (TSadv), and 57.5% (AutoZOOM) for the strongest other attack on each dataset, and has the lowest DTW on BasicMotions and HandMovementDirection (Table 2): (Details: Sec. 4.2, Table 2.)
Other targets and classes. We evaluate transfer to recurrent (LSTM, BiRNN) and convolutional (CNN, TCN, DynamicConv) targets across all 14 datasets, where C2FDA has the best average ASR rank (1.64), ahead of AutoZOOM (2.29), SGAN (5.25), and TSadv (7.86) (Fig. 3). On the same UWaveGestureLibrary class-7/8 test subset, C2FDA reaches 81.0% ASR when its generator excludes those classes (85.7% with all classes), whereas SGAN reaches 28.6% and 71.4%, respectively (Table 13). (Details: Sec. 4.2, Fig. 3; App. D, Table 13.)
References [1] Zerveas et al., A Transformer-based Framework for Multivariate Time Series Representation Learning, KDD 2021.