Table 4 Comparison of recovery performance across frameworks.

From: Hybrid MLOps framework for automated lifecycle management of adaptive phishing detection models

Framework

F1 (Post-recovery)

Recovery latency

\(\Delta\)DP (Post-recovery)

Interpretability retained

HAMF

\(0.9952 \pm 0.0003\)

6 h (auto + HITL)

0.03

\(\checkmark\) (SHAP-guided)

SageMaker19

\(0.9841 \pm 0.0011\)

Manual retrain

0.08

Kubeflow+MLflow16,17

\(0.9863 \pm 0.0009\)

Manual retrain

0.09

\(\blacktriangle\)

PhishBench 2.028,29

\(0.9819 \pm 0.0014\)

Not supported

0.11