Cross-dataset F1-score Matrix for Baseline Intrusion Detection Models

F1-scores ranging from 0.972 to 1.000 for machine learning models evaluated on six intrusion detection datasets. The 5.5 KB Excel file, authored by Hashim Hussain and last updated in May 2026, contains a performance matrix for models like Random Forest, Logistic Regression, and Multi-Layer Perceptron. These models were trained using SMOTE oversampling and 5-fold cross-validation on datasets including NSL-KDD, CIC-IDS2017, and MedBIoT.

Use Cases

• Benchmarking intrusion detection model performance across diverse IoT datasets based on the described F1-score matrix.
• Evaluating the impact of SMOTE oversampling on imbalanced cybersecurity data based on the reported 5–20% F1-score contribution.
• Comparing the latency and detection trade-offs of post-quantum cryptographic components like CRYSTALS-Kyber and SPHINCS+ based on the described 3.1–4.4 second handshake latency.
• Designing adaptive zero-trust policy decisions (ALLOW, MONITOR, DENY, QUARANTINE) based on ML-derived dynamic trust scores mentioned in the description.

Strengths

• Reports high model performance with F1-scores from 0.972 to 1.000 across six datasets.
• Evaluation includes modern IoT-specific traffic from datasets like Edge-IIoTset and IoT-23.
• Model training methodology is detailed, including SMOTE oversampling and 5-fold cross-validation.

Limitations

• The dataset is very small at 5.5 KB, indicating limited scope, likely containing only summary performance metrics.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for certain analyses.

Provenance

Source

Hashim Hussain via figshare

Collection Method

Results from evaluating machine learning classifiers on six public intrusion detection datasets.

Freshness

Last updated 2026-05-28 17:42:35; freshness should be verified.