Worms Dataset for Adversarial Machine Learning Evaluation

A dataset used to evaluate the adversarial robustness of machine learning models against cyber threats. The data, shared by Sudip Saha on figshare, was last updated on June 1, 2026. It is a small dataset of 5.5 KB, stored in an XLS file.

Use Cases

• Benchmarking model robustness against Fast Gradient Sign Method (FGSM) attacks based on the described adversarial training framework.
• Evaluating model performance under Projected Gradient Descent (PGD) attacks as mentioned in the experimental results.
• Training hybrid adversarially-trained models that balance clean accuracy and adversarial loss components.
• Analyzing convergence stability and gradient dynamics during adversarial training as referenced in the description.

Strengths

• Performance metrics are provided, including 97.88% accuracy on clean data and 84.9% accuracy under FGSM attacks.
• The dataset is openly available under a CC-BY-4.0 license.
• Specific adversarial robustness improvements of 6–10 percentage points over CNN and LSTM baselines are documented.

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 at 5.5 KB, indicating limited scope.

Provenance

Source

figshare, author Sudip Saha

Collection Method

Likely generated or compiled for research on adversarial machine learning in cybersecurity.

Freshness

Last updated 2026-06-01 17:55:12; freshness should be verified.