BERT-Embedding-Based JSP Webshell Detection on Bytecode Level Using XGBoost

Table 2

The evaluation metrics results.

Feature extraction	Classification algorithm	Evaluation metrics
		Accuracy	Precision	Recall	F1 score
Word2vec	Bi-LSTM	0.9800	0.9627	0.9685	0.9654
BERT	Bi-LSTM	0.9852	0.9703	0.9789	0.9746
Word2vec	XGBoost	0.9859	0.9805	0.9701	0.9753
BERT	XGBoost	0.9914	0.9868	0.9803	0.9835
Word2vec	RandomForest	0.9844	0.9901	0.9562	0.9729
BERT	RandomForest	0.9842	0.9865	0.9589	0.9724
Word2vec	Support vector machine	0.9852	0.9926	0.9561	0.9740
BERT	Support vector machine	0.9828	0.9652	0.9751	0.9700