OHNN: An Optimized Hybrid Neural Network for Intrusion Detection Systems Using TLBO and CTO

In recent years, cybersecurity has become a hot topic for exploration among researchers. The anomaly-based Detection methods have been widely used for early detection and mitigation of Intrusions. As the hardware is evolving, there is a need for exploration of new features. This is made possible through novel deep learning models. In this paper, an automated construction of various hybrid deep learning models is performed through a twin optimization algorithm. The Teaching Learning Based Optimization and Class Topper Optimization are used to generate new deep learning structures. The model parameters, like the Number of Hidden layers, the Optimizer and the Learning Rate, are controlled by TLBO. CTO was used to select different internal parameters such as Types of layers, Number of nodes per layer, the activation function and initializers. The proposed model was used to explore several deep learning models to train and test on the NSL-KDD dataset. Binary and Multiclass classifiers were explored for different neural network architectures based on TLBO and CTO parameters. The explored models showed comparable accuracy with respect to existing detection models.