Statistical Significance Test Results for Intrusion Detection System Benchmarking

5.5 KB of statistical test results from a framework evaluating machine learning models for network intrusion detection. The dataset, authored by Shailendra Mishra and last updated in April 2026, contains results from Wilcoxon signed-rank, McNemar’s, and DeLong tests applied to models like Random Forest on harmonized NSL-KDD and CICIDS2017 datasets.

Use Cases

• Benchmarking supervised, unsupervised, deep learning, and ensemble models based on the described framework.
• Validating model performance improvements from feature harmonization using the provided statistical test results.
• Comparing the effectiveness of Random Forest against other learning models based on reported accuracy and F1-scores.
• Assessing the proposed cryptographic logging mechanism for experiment reproducibility based on the described SHA-256 hash chaining.

Strengths

• Includes results from multiple statistical validation tests (Wilcoxon signed-rank, McNemar’s, DeLong).
• Reports specific model performance metrics, such as 98.0% accuracy and 97.0% F1-score for Random Forest.
• Framework focuses on reproducibility and transparency, incorporating a tamper-evident logging mechanism.
• Released under a permissive CC-BY-4.0 license.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The manual feature alignment approach described may not be effective for highly heterogeneous datasets.

Provenance

Source

figshare

Collection Method

Results generated from a proposed unified evaluation framework for IDS systems.

Time Range

Experimental timeframe not specified.

Freshness

Last updated 2026-04-20 17:36:52; freshness should be verified.

Geography

Geographic coverage not specified.