LogCAD: An Efficient and Robust Model for Log-Based Conformal Anomaly Detection

Table 1

Comparison between LogRobust and LogCAD.

Method	Learning model	Incremental updating	Robustness	Online
LogRobust	Deep learner (Bi-LSTM with attention)	Semi-retraining by dynamically updating parameters	Semantic vectorization	No
LogCAD	Shallow learner (LR, NB, SVM, etc.) + ensembler + conformal predictor	Updating nonconformity score sets without retraining	Dynamically updated nonconformity score sets	Yes