الفهرس | Only 14 pages are availabe for public view |
Abstract supervisory control and data acquisition (SCADA) is a tool set of software applications for monitoring and controlling industrial processes, that is the collecting of data in real time from remote locations to control equipment and conditions. SCADA supports organizations with the tools which are required to make and deploy data-driven decisions related to their industrial processes. The basic structure of these systems consist of a “master” unit system (generally fully redundant or “fault tolerant”) such as computers or servers that communicates, by utilizing one or more of a multiple of possible telecommunication modes using different communications protocols, to multiple, remote, electronic units (called remote terminal units (RTUs)) which are interacted with the field-based process equipment to gather information and transfer it to the master unit to analyze these data then monitor this information to the operator to make a right decision. The beginning of SCADA systems were isolated networks with no remote interference. SCADA systems network are propagated and interacted with multi vendors to meet the requirements of industrial environments. The data transfers using communications protocols which have a strong point and many advantages but in other hand have vulnerabilities which attackers utilizing this to exploit the industrial systems these attacks have a disaster effect in industrial environments. so, this thesis introduces the vulnerabilities of the communications protocols and how the attacker exploits this to destroy system then introduce multiple techniques that used to detect the behavior of the attacks comparing the normality of the industrial system especially using deep learning neural networks using simulation of dataset and applies ensemble algorithm with decision tree (DT) and support vector machine (SVM) multiple classification in this dataset using Python 3.0 programming language and Google Collab editor to simulate this two classifications models in the dataset, and compare the results between them with known characteristics such as accuracy , Precision , recall ,and F1 score. And comprises these results with Long-Short Term Memory (LSTM) algorithm. This study could be applied in the industrial environment to achieve the benefit to the industrial application for protecting them from the electronic attacks. |