CICIDS2017 Ablation Study Results for Healthcare IoT Intrusion Detection

Ali Alqazzaz published ablation study results on the CICIDS2017 benchmark dataset on April 17, 2026. The 5.5 KB Excel file contains results from evaluating the MedDefender-MHAN explainable multi-head attention network for intrusion detection in healthcare IoT environments. The model achieved a reported detection accuracy of 99.47% on CICIDS2017.

Use Cases

• Benchmarking intrusion detection models based on reported accuracy and latency metrics.
• Studying the impact of model architecture changes (ablation) on performance for healthcare IoT security.
• Comparing explainability methods for AI-driven security based on alignment with expert-annotated attack signatures.
• Evaluating real-time inference capabilities for clinical deployment based on sub-3ms latency and throughput figures.

Strengths

• Specific performance metrics are provided, including 99.47% detection accuracy on CICIDS2017.
• Includes explainability evaluation metrics: 94.6% alignment with expert signatures and 91.9% temporal accuracy.
• Real-time performance is quantified with sub-3ms inference latency and 435 samples/second throughput.

Limitations

• Row count and column-level documentation are absent; field semantics must be inferred after download.
• The dataset is very small (5.5 KB), suggesting it contains summary results rather than raw training or test data.

Provenance

Source

figshare, author Ali Alqazzaz.

Collection Method

Results from an ablation study on the CICIDS2017 benchmark dataset.

Time Range

The study was published in 2026; the temporal coverage of the underlying CICIDS2017 data is not specified.

Freshness

Last updated 2026-04-17 17:39:40; freshness should be verified.

Geography

Geographic coverage is not specified.