The new era of data networking involves Deep Neural Network for more efficiency and productive output. An intelligent autonomous intrusion detection system alone is not enough, when data security is important. Explainable Artificial Intelligence is based on standards and do not bypass human interference to make decision on intrusion flags. Ensuring network security is crucial and essential in Oil and Gas industries, especially for remote sites such as Wellheads, Gas Gathering Manifold and Remote Headers. Data transmission to plant and further to corporate network from remote sites is vulnerable and a target for attackers due to remoteness of the sites. These sites are mostly unmanned and remotely being monitored and controlled over network. This paper is presenting a review, analysis of integrating Explainable Artificial Intelligence (XAI) with Deep Neural Network (DNN) for Intrusion Detection System of IIoT’s data transmission to corporate network in Oil and Gas industries. Here, we tried to explore and research about recent advancements in field of deep neural network-based intrusion detection system. This paper is presenting the idea to implement XAI integrated DNN for intrusion detection system in oil and Gas industries to protect any kind of cyber-attack on processed or raw data from remote sites. Many research papers have been analyzed and studied during the research and found many gaps which need to be filled when developing a smart intrusion detection system. Traditional black box transparent theory is not enough to combat cyberattacks but it needs human interface to have explanations of flags. Deep Neural Network based learning, training and testing made this research paper more accurate for intrusion detection when working with critical process data coming from IIoTs of unmanned sites of Oil and Gas industries. Weight based feature selection and attack analysis based on explanations based neural networks and pre-defined standards making this intrusion detection system best for complex networks of IIoTs.
Haque et al. (Wed,) studied this question.