Performance Metrics and Efficiency Trade-offs for Intrusion Detection Systems

A 5.5 KB Excel dataset presents results from a unified framework for evaluating machine learning-based Intrusion Detection Systems (IDS). The framework harmonizes features from the NSL-KDD and CICIDS2017 datasets and benchmarks models including Random Forest, which achieved 98.0% accuracy and 97.0% F1-score. Authored by Shailendra Mishra and last updated on April 20, 2026, this work focuses on reproducibility and statistical validation in cybersecurity research.

Use Cases

• Benchmarking supervised, unsupervised, deep learning, and ensemble models for IDS based on the described harmonized datasets.
• Analyzing the impact of feature harmonization on model performance using the described ablation analysis.
• Comparing statistical validation methods for IDS evaluations, referencing the Wilcoxon signed-rank, McNemar’s, and DeLong tests mentioned.
• Investigating cryptographic logging mechanisms for experiment reproducibility, as described with the SHA-256 hash chaining approach.

Strengths

• Includes specific performance metrics for a leading model, such as 98.0% accuracy and 97.0% F1-score.
• Proposes a framework integrating feature harmonization, multi-model benchmarking, and statistical validation for improved transparency.
• Released under a permissive CC-BY-4.0 license, facilitating reuse and sharing.

Limitations

• The dataset is very small at 5.5 KB, suggesting it contains summary results or metadata rather than raw network traffic data.
• Row count and column-level documentation are unknown, limiting assessment of the underlying data structure.
• The description notes the feature harmonization approach is manual and might not be effective for highly heterogeneous datasets.

Provenance

Source

figshare

Collection Method

Results from an experimental framework harmonizing and benchmarking on the NSL-KDD and CICIDS2017 datasets.

Time Range

null

Freshness

Last updated 2026-04-20 17:36:55

Geography

null