Intelligent intrusion detection system in smart grid using computational intelligence and machine learning

Abstract

Smart grid systems enhanced the capability of traditional power networks while being vulnerable to different types of cyber-attacks. These vulnerabilities could cause attackers to crash into the network breaching the integrity and confidentiality of the smart grid systems. Therefore, an intrusion detection system (IDS) becomes an important way to provide a secure and reliable services in a smart grid environment. This article proposes a feature-based IDS for smart grid systems. The proposed system performance is evaluated in terms of accuracy, intrusion detection rate (DR), and false alarm rate (FAR). The obtained results show that the random forest and neural network classifiers have outperformed other classifiers. We have achieved a 0.5% FAR on KDD99 dataset and a 0.08% FAR on the NSLKDD dataset. The DR and the testing accuracy on average are 99% for both datasets. INTRODUCTIONData-driven technologies is now applied to smart grid as a way of sustainable energy environment. This approach can be added to a cyber-physical system consisting of hardware, software, and other physical gears. Smart grid supplies electricity on-demand to end-users from centralized stations and distribute to generating stations using information and communication technologies. Energy supplier companies supply electricity at low cost and also control the end-user demand for supply. In the smart grid system, one of the significant issues is security. Many vulnerabilities exist in cyber-physical systems and hackers take advantage of vulnerabilities to launch malicious attacks on power systems. Security problems usually include authentication, data protection, availability, confidentiality, honesty, energy efficiency, single-point failures to be tested, and so on. 1 The attackers destroy a whole range of cyberspace in modern electronic warfare. In our societies cybercrimes proliferated. Attacks, hacking, and malicious practices such as viruses, trojans, and spamming are common risks to individuals and nations. The digital networks of cellular telephony, wireless sensor networks, satellites, tactical military communications, Internet of things, smart grids and supervisory control, and data acquisition are everything vulnerable to that kind of electronic attack. 2 A lot of work has been done on smart grid system implementation but the majority of work are not focusing on the security requirements for the smart grid systems. 3,4 Intrusion detection system (IDS) plays an essential role in cyber-attacks on smart grid systems and secures them against attacks. The IDS are part of the network security domain and play a vital role in protecting and maintaining a secure network. IDS system is represented in Figure 2. A typical IDS system examines and analyzes network traffic to detect and analyze attacks, and also to prevent any security violations by generating alarms for network administrator. There are two major types of IDS: host-based IDS and

An Intelligent Intrusion Detection System in Smart Grid Using PRNN Classifier

“…These performance metrics are based on the True positive (TP), True Negative (TN), False positive (FP), and False Negative (FN). The analysis is compared with existing techniques to validate proposed system efficiency [21]. Tab.…”

Adaptive Machine Learning Based Distributed Denial-of-Services Attacks Detection and Mitigation System for SDN-Enabled IoT

“…Many advanced solutions provide successful intrusion detection with the support of machine learning. Such solutions are widely deployed for different applications such as auditing applications and smart grid and IoT [ 31 , 32 , 33 ] Some of the solutions provide virtual machine learning-based DDoS detection systems for SDN-enabled networking, but still demand significant enhancements to develop industrial level security solution [ 28 ].…”

Analysis of Autoencoders for Network Intrusion Detection

“…The signature-based ML-NIDS method generally assumes that pre-labeled attack traffic data are made available in advance, and learns those known intrusion patterns using supervised learning algorithms [ 7 , 8 ], such as SVM [ 9 , 10 ] and Decision Tree [ 11 ]. Khan et al [ 12 ] used a particle swarm optimization method to select the optimal feature subset from a given dataset in terms of detection accuracy. The authors then evaluated several machine-learning algorithms such as random forest and neural network as classifiers using KDD Cup 99 [ 13 ] and NSL-KDD [ 14 ] datasets.…”