Hybrid Security Framework for Critical Systems: Combining Attack Graph and Intrusion Detection, Deep Learning

Security modeling and graphical security are popular methods for assessing a system’s security. These models reflect security situations and assist security specialists in identifying the system’s vulnerabilities. One common technique for mathematically and graphically illustrating the events that culminate in a successful cyber-attack is using attack graphs. Network complexity and its many platforms and diverse software products resulted in multiple levels of connectivity, and this complexity of environments offers a severe issue for cybersecurity and overall organizational security. An efficient way to guarantee computer network security is through intrusion detection systems, which can detect anomalous data in intricate network environments. The structural and dynamic complexity of critical systems cannot be captured by the traditional methods of system vulnerability and risk analysis; cyber-attacks have become more planned and precise and have clear goals targeting weaknesses in systems’ security. This study uses an attack modeling technique for critical systems, specifically the attack graph technique. The research seeks to present a hybrid security framework by merging attack graphs with intrusion detection systems and deep learning to increase the security of critical systems and lead to reducing vulnerability in system networks while boosting the detection power of intrusion detection systems. A case study was conducted on the petroleum SCADA system. Security data for the SCADA system has been identified, as have major weaknesses, and an attack graph is presented.