Table 1.

Comparison of the literature.

Reference Number	Method	Drawback	ACC
[16]	Combined EA, SVM, and ANN	Limited dataset	99.3%
[18]	Hybrid-based IDS	Fixed set of features	96.64%
[19]	ML IDS for MIoT	Single dataset used	99.9%
[21]	Dynamic Anomaly Detection Scheme	Only on AODV-based	84.0%
[24]	Mix machine learning techniques	Small dataset, fixed features	99.5%
[25]	Combined DTF, CNN, and LSTM	Only wormhole detection	96%
[26]	Web-based DDoS detection	Only web single dataset	99%
[27]	Mining sequences of IP’s	Some worst performance	-
[28]	Building and evaluation using ANN-MLP	Single dataset: UNSW-NB15	76.96%
[29]	Detection by ensemble of neural classifiers	Overfitting	99.4%
[30]	Detection by MLP, NB, and RF	Not applicable for all attacks	98.63
[35]	Detection by CNN and LSTM	Not applicable for low volumes	96.7