A Deep Learning-Based Framework for Feature Extraction and Classification of Intrusion Detection in Networks

Table 3

Comparison of different feature selection techniques with the proposed feature selection model.

Method	Feature selection	Dataset	Features used	Accuracy
Ahmadi et al. [36]	Chi squared, information gain, correlation-based evaluation	NSL-KDD	20	80.6
Liu et al. [37]	ANOVA	NSL-KDD, UNSWNB15, CICIDS-2017	16, 13, 39	83.28
Gottwalt et al. [39]	CorrCorr	NSL-KDD, UNSWNB15	19	95
Vinutha et al. [40]	Chi squared, information gain, gain ratio, correlation-based attribute evaluation, symmetrical uncertainty	NSL-KDD	31	85.91
Tang et al. [41]	SDN environment-based six basic features	NSL-KDD	6	75.75
Bhattacharya et al. [42]	Layered wrapper feature selection approach	NSL-KDD	16	83.14
Rama et al. [43]	Hyper graph-based genetic algorithm (HG-GA)	NSL-KDD, KDD-99	35	97.14
Mohammadi et al. [44]	Linear correlation, cuttlefish algorithm, decision tree	KDD99	10	95.03
Gao et al. [45]	CART algorithm	NSL-KDD	17	79.7
Tang et al. [46]	Stacked autoencoder	NSL-KDD	Autoselection	87.74
Proposed model	ANOVA, chi square, PCA	NSL-KDD	27	99.73