Towards Effective Detection of Recent DDoS Attacks: A Deep Learning Approach

Table 6

Classification results achieved by each feature selection strategy within ensemble group.

Ensemble group	Feature selection strategy	Accuracy	Precision	Recall	F1-Score
Ensemble group 1	Threshold-based list	0.9319	0.9355	0.9315	0.9150
	Top 10 Weight	0.9954	0.9956	0.9956	0.9950
	Threshold-based list + Top 10 Weight	0.9950	0.9950	0.9950	0.9950
Ensemble group 2	Threshold-based list	0.9519	0.9540	0.9520	0.9515
	Top 10 Weight	0.9962	0.9965	0.9965	0.9960
	Threshold-based list + Top 10 Weight	0.9941	0.9940	0.9940	0.9940
Ensemble group 3	Threshold-based list	0.9198	0.9200	0.9200	0.9195
	Top 10 Weight	0.9949	0.9950	0.9950	0.9950
	Threshold-based list + Top 10 Weight	0.9910	0.9910	0.9910	0.9910
Ensemble group 4	Threshold-based list	0.9277	0.9325	0.9275	0.9275
	Top 10 Weight	0.9877	0.9880	0.9880	0.9880
	Threshold-based list + Top 10 Weight	0.9940	0.9950	0.9950	0.9940

Receiver operating characteristic/area under the curve (ROC/AUC).