Ramadan [
2] proposed a hybrid IDS system where a pre-processing phase is utilized to reduce the required time. The feature selection process is done by using the Enhanced Shuffled Frog Leaping (ESFL) algorithm, and the selected features are classified using the Light Convolutional Neural Network with Gated Recurrent Neural Network (LCNN-GRNN) algorithm. Maha [
3] Designed an intelligent BBFO-GRU instrusion detection systems in Industrial Cyber-Physical environment based on the Gated Recurrent Unit (GRU) model. In addition, in order to enchance the detection rate, NADAM optimizer is utilized to optimize the GRU hyperparameters. Derhab [
4] designed a Temporal Convolution Neural Network (TCNN) in IoT, which combines the Convolution Neural Network (CNN) with a causal convolution. TCNN with Synthetic Minority Oversampling Technique-Nominal Continuous (SMOTE-NC) is evaluated on Bot-IoT dataset. Mulyanto [
5] implemented a cost-sensitive neural network based on focal loss, called the focal loss network intrusion detection system (FL-NIDS), in order to overcome the problem of imbalanced data . FL-NIDS was applied using DNN and convolutional neural network (CNN). To evaluate this approach, three benchmark intrusion detection datasets that suffer from imbalanced distributions were used: NSL-KDD, UNSW-NB15, and Bot-IoT. Azmin [
6] proposed a new paradigm of the synthesizing task based on Variational Laplace AutoEncoder (VLAE), and Deep Neural Network (DNN) classifier. The authors evaluated the model on the NSL-KDD dataset. Jie [
7] proposed an Intrusion Detection System based on bidirectional simple recurrent unit. In addition, the skip connections is used to to alleviate the vanishing gradient problem and improve the training effectiveness. Mahboob [
8] employed the butterfly optimization algorithm (BOA), and meta-heuristic to perform feature selection. A multilayer perceptron (MLP) classifier was used to evaluate the capability of the selected features to predict attacks. In addition to the gradient descent (GD) training method, two other metaheuristic methods, particle swarm optimization (PSO) and genetic algorithm (GA) were used to optimize the classification structure. This approach was tested on the NSL-KDD dataset. Sahar [
9] developed a network intrusion detection system based on deep learning, and implemented in the fog node for attack detection. The datasets used are UNSW-NB15 and NSL-KDD. Khan [
10] conceived an intrusion detection system, based on convolutional neural network algorithm. The entire network consists of three hidden layers. Each hidden layer contains a convolutional layer and a pooling layer. Bediya [
11] discussed many possible attacks at IoT networks and distributed denial of service (DDoS) attack. Then, the author proposed a blockchain-based IDS for the IoT network, called BIoTIDS. Khan [
12] implemented a convolutional recurrent neural network (CRNN) to create a DL-based hybrid ID framework that predicts and classifies malicious cyberattacks in the network. In the HCRNNIDS, the convolutional neural network (CNN) performs the convolution to capture local features, and the recurrent neural network (RNN) captures temporal features to improve the ID system’s performance and prediction. Experiments were carried out on the CSE-CIC-DS2018 dataset. Soumyadeep [
13] presented an unique Generic-Specific autoencoder model where the generic one learns the features that are common across all forms of network intrusions, and the specific ones learn features that are pertaining only to that domain. Sekhar [
14] applied a deep Autoencoder with Fruitfly Optimization. Firstly, the missing values in the dataset have been imputed with the Fuzzy C-Means Rough Parameter (FCMRP) algorithm, which handles the imprecision in datasets with the exploit of fuzzy and rough sets while preserving crucial information. Then, robust features are extracted from the Autoencoder with multiple hidden layers. Finally, the obtained features are fed to the backpropagation neural network (BPN) to classify the attacks. Experiments have been carried out on the NSL-KDD and UNSW-NB15 dataset. Khonde [
15] proposed a hybrid method, based on semi-supervised machine learning classifiers. Moreover, classifiers used are Support vector machine, decision tree and k-nearest neighbor. Experiments were conducted on NSL-KDD dataset. Shen [
16] proposed an ensemble method, combining the extreme learning machine (ELM) as a base classifier, and a pruning method based on the Bat Algorithm (BA) as an optimizer. Deepa [
17] used the K-Means Algorithm features. Moreover, authors combined Cuckoo Search Optimization (CSO) and the K-Means clustering algorithm. This approach was tested on different datasets. Divakar [
18] used an ensemble method based on XGB Classifier on UNSW-NB 15 dataset.