An efficient evolutionary deep learning-based attack prediction in supply chain management systems

Abstract

Supply Chain Management Systems (SCM) is the critical infrastructure that can be treated as a significant factor since it forms advancement in intelligent devices. The cyber-attacks against SCM are increasing day by day, and detecting these cyber-related attacks remains a challenging task. In general, deep learning algorithms are superior in handling complex datasets than machine learning approaches because of their layered structure and practical algorithms to extract the relevant information from the input training data. Deep Learning (DL) methodologies are rapidly being utilized to counter cyber-attacks on Supply Chain Management (SCM) systems. This research proposes a new method that leverages evolutionary and DL approaches to detect cyber-attacks in a cloud-based SCM environment. The input data undergoes preprocessing to convert it to a suitable format. To reduce the complexity of high dimensional data and select the most relevant features, Evolution Social Spider Optimization (ESSO) is employed. A Deep Belief Network (DBM) trained by an Extreme Learning Machine (ELM) is then introduced to identify and classify cyber-attacks. The performance of the proposed network is improved through fine-tuning using Poor and Rich Optimization (PRO) algorithms. The proposed PRO-optimized ELM-trained DBM model aims to enhance security in the SCM environment by recognizing and classifying intrusions. The performance of the proposed model is experimented on benchmark datasets and compared with traditional approaches. The depicted results based on various measures show that the proposed ELMDBN-PRO in Blockchain of Things-Internet of Things (BoT-IoT) dataset achieves an average accuracy of 99%, mod bus dataset achieves 99.5% over recent approaches.