Ablation Study of a Hybrid Risk-Based Authentication Framework

K. Sasikumar's study proposes a hybrid Risk-Based Authentication framework integrating ensemble classifiers, fuzzy logic, and optimization. Experimental results report 97.77% accuracy and 98.72% F1-score for detecting account compromise based on login behavior, device, and network data. The dataset, last updated in May 2026, is a 5.5 KB Excel file containing the study's results.

Use Cases

• Benchmarking ensemble models for account takeover detection based on the described Gradient Boosting, SVM, and XGBoost combination.
• Analyzing the impact of fuzzy logic and L-BFGS-B optimization on risk threshold refinement for authentication decisions.
• Studying the application of SHAP for explainable AI in cybersecurity risk assessment models.
• Evaluating authentication framework performance on large-scale datasets based on the reported metrics for 2M to 30M records.

Strengths

• Reports specific performance metrics including 97.77% accuracy and a 0.0303 Equal Error Rate.
• Describes testing on large-scale datasets ranging from 2 million to 30 million records.
• Framework integrates multiple techniques: ensemble learning, fuzzy logic, clustering, and optimization.

Limitations

• Dataset is very small (5.5 KB), suggesting it contains summary results rather than raw behavioral data.
• Row count and specific column definitions are unknown, limiting direct usability.
• Description metadata is limited; actual data quality and structure require manual inspection after download.

Provenance

Source

figshare, author K. Sasikumar.

Collection Method

Likely contains experimental results from a proposed hybrid RBA framework.

Freshness

Last updated 2026-05-26 17:44:55; freshness should be verified.