Empirical Evaluation of Noise Influence on Supervised Machine Learning Algorithms Using Intrusion Detection Datasets

Table 3

The total number of instances before and after excluding the outliers and the extreme values.

Dataset	Data file	Before filtering			After filtering
		Total no. of instances	No. of benign instances	No. of attack instances	Total no. of instances	No. of benign instances	No. of attack instances
NSL-KDD	KDDTrain+	125,973	67,343	58,630	45,331	12,360	32,971
	KDDTest+	22,544	9,711	12,833	2,670	2,149	521
UNSW-NB15	UNSW-NB15 training	175,341	56,000	119,341	54,729	7,503	47,226
	UNSW-NB15 testing	82,332	37,000	45,332	31,468	10,045	21,423