# Image Classification on CIFAR-10 (test) (Specific Adversarial Attack Robustness)

Source: [Image Classification on CIFAR-10 (test) (Specific Adversarial Attack Robustness)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-test-specific-adversarial-attack-robustness>)

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

Data updated: 2026-02-26T05:34:21.383Z

Evaluation of Image Classification performance on the CIFAR-10 test set, specifically measuring accuracy under various adversarial perturbation attacks, including Label Flipping, Gaussian Noise, Sign Flipping, and Scaling attacks.

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 | Image Classification |
| Dataset (including any stated split) | CIFAR-10 (test) |
| Metric | Accuracy (No Attack) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 5 of 5 |
| 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

- [Accuracy (No Attack)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-test-specific-adversarial-attack-robustness.md?metric=Accuracy%20(No%20Attack)>): Higher is better; unit: PERCENT.

- [Accuracy (Label Flipping Attack)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-test-specific-adversarial-attack-robustness.md?metric=Accuracy%20(Label%20Flipping%20Attack)>): Higher is better; unit: PERCENT.

- [Accuracy (Gaussian Noise Attack)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-test-specific-adversarial-attack-robustness.md?metric=Accuracy%20(Gaussian%20Noise%20Attack)>): Higher is better; unit: PERCENT.

- [Accuracy (Sign Flipping Attack)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-test-specific-adversarial-attack-robustness.md?metric=Accuracy%20(Sign%20Flipping%20Attack)>): Higher is better; unit: PERCENT.

- [Accuracy (Scaling Attack)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-test-specific-adversarial-attack-robustness.md?metric=Accuracy%20(Scaling%20Attack)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy (No Attack) | Result date | Report paper |
| --- | --- | --- | --- |
| Cost-TrustFL | 91.2 | Not recorded | 694b737653fdd5c36173996f |
| FLTrust | 90.5 | Not recorded | 694b737653fdd5c36173996f |
| Trimmed-Mean | 89.8 | Not recorded | 694b737653fdd5c36173996f |
| FedAvg | 89.1 | Not recorded | 694b737653fdd5c36173996f |
| Krum | 88.5 | Not recorded | 694b737653fdd5c36173996f |

### Cost-TrustFL

Accuracy (No Attack): 91.2. Result ID: 699944465698665ec4ae6221.

[Reporting paper](<https://www.sota2.com/research/paper/694b737653fdd5c36173996f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/694b737653fdd5c36173996f.md>)

Recorded evaluation configuration: {"Malicious client ratio":"30%","Dirichlet alpha (α)":"0.5","Backbone":"CNN (2 conv, 2 fc layers)","Communication rounds":"200","Local epochs (E)":"5","Batch size":"32","Learning rate":"0.01","Reference dataset size":"100 samples","Cost weight (λ)":"0.3"}. This record does not assert independent reproduction.

### FLTrust

Accuracy (No Attack): 90.5. Result ID: 699944465698665ec4ae6220.

[Reporting paper](<https://www.sota2.com/research/paper/694b737653fdd5c36173996f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c451e209131cea3325c.md>)

Recorded evaluation configuration: {"Malicious client ratio":"30%","Dirichlet alpha (α)":"0.5","Backbone":"CNN (2 conv, 2 fc layers)","Communication rounds":"200","Local epochs (E)":"5","Batch size":"32","Learning rate":"0.01","Reference dataset size":"100 samples"}. This record does not assert independent reproduction.

### Trimmed-Mean

Accuracy (No Attack): 89.8. Result ID: 699944465698665ec4ae621f.

[Reporting paper](<https://www.sota2.com/research/paper/694b737653fdd5c36173996f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69531e209131cea218fb.md>)

Recorded evaluation configuration: {"Malicious client ratio":"30%","Dirichlet alpha (α)":"0.5","Backbone":"CNN (2 conv, 2 fc layers)","Communication rounds":"200","Local epochs (E)":"5","Batch size":"32","Learning rate":"0.01"}. This record does not assert independent reproduction.

### FedAvg

Accuracy (No Attack): 89.1. Result ID: 699944465698665ec4ae621d.

[Reporting paper](<https://www.sota2.com/research/paper/694b737653fdd5c36173996f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e761e209131cea40072.md>)

Recorded evaluation configuration: {"Malicious client ratio":"30%","Dirichlet alpha (α)":"0.5","Backbone":"CNN (2 conv, 2 fc layers)","Communication rounds":"200","Local epochs (E)":"5","Batch size":"32","Learning rate":"0.01"}. This record does not assert independent reproduction.

### Krum

Accuracy (No Attack): 88.5. Result ID: 699944465698665ec4ae621e.

[Reporting paper](<https://www.sota2.com/research/paper/694b737653fdd5c36173996f.md>)

Recorded evaluation configuration: {"Malicious client ratio":"30%","Dirichlet alpha (α)":"0.5","Backbone":"CNN (2 conv, 2 fc layers)","Communication rounds":"200","Local epochs (E)":"5","Batch size":"32","Learning rate":"0.01"}. This record does not assert independent reproduction.

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