Adversarial Robustness Evaluation Results for a Hybrid Deep Learning Classifier

Sudip Saha published a dataset on figshare in 2026 detailing the performance of a novel adversarially-trained classifier. The data likely contains experimental results, including accuracy metrics on clean and adversarial data, training dynamics, and calibration errors. The dataset is small, at 5.5 KB, and is available in XLS format.

Use Cases

• Benchmarking adversarial robustness based on reported accuracy under FGSM and PGD attacks
• Analyzing training convergence stability based on mentioned gradient norm decay and runtime efficiency metrics
• Studying the trade-off between clean accuracy and adversarial robustness based on the described dual-loss optimization

Strengths

• Performance metrics are quantified, including 97.88% clean accuracy and 84.9% accuracy under FGSM attacks
• Results are compared against established baselines (CNN, LSTM) with a reported 6–10 percentage point improvement
• Includes specific metrics like a 30% reduction in expected calibration error

Limitations

• Row count is unknown, which may limit suitability assessment
• Column-level documentation is absent; field semantics must be inferred after download
• The dataset is very small (5.5 KB), indicating limited scope, likely containing summary results rather than raw training data

Provenance

Source

figshare

Collection Method

Likely contains results from experiments on reconnaissance, shellcode, and worms datasets as described in the associated paper.

Freshness

Last updated 2026-06-01 17:55:08; freshness should be verified