Table 1.

Confusion Matrices.

Model	TP	FN	FP	TN	Accuracy Trend	Observations
BERT	8742	59	42	958	Highest accuracy and balance	Best at detecting both classes with low error rate.
LSTM	8650	151	134	866	High accuracy, balanced performance	Handles both classes better; ideal for real-time detection.
SVM	8604	197	180	820	Moderate accuracy, high FP	Overpredicts DoS as BENIGN; moderate recall but low precision.
Naïve Bayes	8350	451	385	615	High FN & FP	Fails to distinguish BENIGN and DoS due to data imbalance.