Detekcija napada na senzorske signale u industrijskim kontrolnim sistemima

Abstract

U cilju povećanja produktivnosti i efikasnosti proizvodnje, četvrta industrijska revolucija vodi ka implementaciji kibernetsko fizičkih sistema i interneta stvari u industrijskom okruženju. Sveobuhvatna komunikacija čini kibernetsko fizičke sisteme podložnim na spoljašnje uticaje, koji često mogu imati negativnu nameru, npr. napadi i smetnje proistekli od različitih uzročnika. Uticaj napada na sistem može dovesti do anomalija i ozbiljnih posledica po delove sistema ili sistem u celosti. Stoga, odbrambeni mehanizmi za pravovremenu detekciju napada moraju biti razvijeni, kako bi se sistem zaštitio i održala njegova funkcionalnost. U ovom radu, predložen je metod za detekciju napada na senzorske signale u kontinualno upravljanim sistemima. Metod je baziran na mašinama sa nosećim vektorima, a testiran na skupu podataka iz sistema za preradu vode.

To improve productivity and efficiency in industrial manufacturing, the fourth industrial revolution leads to the implementation of Cyber Physical Systems (CPS) and Internet of Things (IoT) in the industrial environment. Ubiquitous communication makes CPS susceptible to external influences, which can have a negative intention; for instance, CPS are prone to various attacks and malicious threats by different adversaries. The impact of an attack on the system can lead to anomalies and serious consequences for system parts or the system as a whole. Security mechanisms must be developed in order to timely detect different attacks and to keep the system safe and protected. In this paper, a method for sensor signal attacks detection in a continuous time controlled systems has been proposed. The method is based on Support Vector Machines (SVM) and tested on the data obtained from the Secure Water Treatment (SWaT) testbed, a scaled-down plant that produces purified water.