Novel Hybrid Model for Intrusion Prediction on Cyber Physical Systems' Communication Networks based on Bio-inspired Deep Neural Network Structure

There are growing concerns on the security of communication networks of Cyber Physical Systems (CPSs). In a typical Cyber Physical System (CPS), the plant, actuators, sensors and controller interface through a communication network, which enable computing and data transmission in the CPS. Consequently, the communication network is vulnerable to sophisticated attacks. Attacks on CPSs communication networks can cause damage to critical resources and infrastructure. In this sense, it is crucial to accurately predict these attacks in order to minimise their impact on the target CPSs networks. In this paper, we propose a novel hybrid approach for intrusion prediction on CPSs communication networks. We use a bio-inspired hyperparameter search technique to generate an improved deep neural network structure based on the core hyperparameters of a neural network. Furthermore, we derive a prediction model based on the improved neural network structure and evaluate its performance using two well-known datasets, namely, the CICIDS2017 and UNSW-NB15 datasets. Results obtained from rigorous experimentation show that our model can predict diverse attack types with high accuracy, low error and false positive rates, and outperforms state-of-the-art comparative models.