# Adversarial Defense on ImageNet-1K 1.0 (test)

Source: [Adversarial Defense on ImageNet-1K 1.0 (test)](<https://www.sota2.com/research/sota/adversarial-defense-on-imagenet-1k-1-0-test>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-30T07:36:41.362Z

Evaluation of model robustness and defense mechanisms against adversarial examples on the test split of the ImageNet-1000 dataset, reporting metrics like Classification Accuracy (CA), Perturbation Accuracy (PA), AUC, and Attack Success Rate (ASR).

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Adversarial Defense |
| Dataset (including any stated split) | ImageNet-1K 1.0 (test) |
| Metric | Classification Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 7 of 7 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Classification Accuracy](<https://www.sota2.com/research/sota/adversarial-defense-on-imagenet-1k-1-0-test.md?metric=Classification%20Accuracy>): Higher is better; unit: PERCENT.

- [Perturbation Accuracy](<https://www.sota2.com/research/sota/adversarial-defense-on-imagenet-1k-1-0-test.md?metric=Perturbation%20Accuracy>): Higher is better; unit: PERCENT.

- [AUC](<https://www.sota2.com/research/sota/adversarial-defense-on-imagenet-1k-1-0-test.md?metric=AUC>): Higher is better; unit: SCALAR.

- [Attack Success Rate](<https://www.sota2.com/research/sota/adversarial-defense-on-imagenet-1k-1-0-test.md?metric=Attack%20Success%20Rate>): Lower is better; unit: PERCENT.

## Recorded results

| Method | Classification Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| Platonic Defense | 78.6 | Not recorded | 6a434d7dba73165dbb74395b |
| No Defense | 78.3 | Not recorded | 6a434d7dba73165dbb74395b |
| Decomp | 78.22 | Not recorded | 6a434d7dba73165dbb74395b |
| ZIP | 67.99 | Not recorded | 6a434d7dba73165dbb74395b |
| DeDe | 46.78 | Not recorded | 6a434d7dba73165dbb74395b |

### Platonic Defense

Classification Accuracy: 78.6. Result ID: 6a4371eefbc373592c93ad41.

[Reporting paper](<https://www.sota2.com/research/paper/6a434d7dba73165dbb74395b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a434d7dba73165dbb74395b.md>)

Recorded evaluation configuration: {"Evaluation protocol":"linear probing","Attack":"Average (Adversarial Attacks)"}. This record does not assert independent reproduction.

### No Defense

Classification Accuracy: 78.3. Result ID: 6a4371eefbc373592c93ad3b.

[Reporting paper](<https://www.sota2.com/research/paper/6a434d7dba73165dbb74395b.md>)

Recorded evaluation configuration: {"Evaluation protocol":"linear probing","Attack":"Average (Adversarial Attacks)"}. This record does not assert independent reproduction.

### Decomp

Classification Accuracy: 78.22. Result ID: 6a4371eefbc373592c93ad3c.

[Reporting paper](<https://www.sota2.com/research/paper/6a434d7dba73165dbb74395b.md>)

Recorded evaluation configuration: {"Evaluation protocol":"linear probing","Attack":"Average (Adversarial Attacks)"}. This record does not assert independent reproduction.

### ZIP

Classification Accuracy: 67.99. Result ID: 6a4371eefbc373592c93ad3e.

[Reporting paper](<https://www.sota2.com/research/paper/6a434d7dba73165dbb74395b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71021e209131cea4e7cd.md>)

[Implementation](<https://github.com/sycny/ZIP>)

Recorded evaluation configuration: {"Evaluation protocol":"linear probing","Attack":"Average (Adversarial Attacks)"}. This record does not assert independent reproduction.

### DeDe

Classification Accuracy: 46.78. Result ID: 6a4371eefbc373592c93ad3d.

[Reporting paper](<https://www.sota2.com/research/paper/6a434d7dba73165dbb74395b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777e1e209131cea738c3.md>)

[Implementation](<https://github.com/tardisblue9/DeDe>)

Recorded evaluation configuration: {"Evaluation protocol":"linear probing","Attack":"Average (Adversarial Attacks)"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.