Intrusion Detection System for Internet of Things Based on Temporal Convolution Neural Network and Efficient Feature Engineering

Table 11

Comparison with related work tested on Bot-IoT dataset.

Ref	Model	Task	Accuracy	Precision	Recall	F1-score	Training time (s)
[49]	RNN	Binary	99.7404%	99.9904%	99.7499%	—	8035
	LSTM	Binary	99.7419%	99.9910%	99.7508%	—	10482.19
[61]	Ensemble learning	Binary	99.97%	—	—	—	—
[25]	RNN with BPTT	Multiclass	99.912%	—	—	—	2012
[50]	DeepDCA	Multiclass	98.73%	99.17%	98.36%	98.77%	—
[51]	ANN	Normal/DDoS	100%	100%	100%	100%	—
[52]	FNN	Multiclass	99.02%	—	—	—	—
Our	TCNN	Multiclass	99.9986%	99.9974%	97.4975%	98.6641%	424
Work	LSTM		99.9654%	99.9443%	84.5703%	89.3016%	762
	CNN		99.9973%	95.1360%	97.0783%	96.0500%	419
	LR		99.2858%	74.5496%	98.6987%	78.9640%	709
	RF		97.4586%	77.8298%	98.8643%	84.4592%	191