Interval forecasting of cyber-attacks on industrial control systems

International Conference on Mechanical Engineering, Automation and Control Systems 2017

机械制造;无线电电子学;计算机科学

Ivanyo, Y.M.^1 ; Krakovsky, Y.M.^2 ; Luzgin, A.N.^3

Irkutsk State Agrarian University Named after A.A. Ezhevsky, Molodezhny Settlement, Irkutsk Region Irkutsk

664038, Russia^1

Irkutsk State University of Railway Transport, Chernyshevsky St. 15, Irkutsk

664074, Russia^2

Administration of Irkutsk City, Lenina St. 14, Irkutsk

660025, Russia^3

关键词: Environmental disasters;    Industrial control systems;    Interval forecasting;    Probabilistic neural networks;    Probability estimate;    Protection methods;    Smoothing parameter;    Statistical characteristics;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/327/2/022044/pdf
DOI  :  10.1088/1757-899X/327/2/022044

学科分类：计算机科学（综合）

来源: IOP

【 摘 要 】

At present, cyber-security issues of industrial control systems occupy one of the key niches in a state system of planning and management Functional disruption of these systems via cyber-attacks may lead to emergencies related to loss of life, environmental disasters, major financial and economic damage, or disrupted activities of cities and settlements. There is then an urgent need to develop protection methods against cyber-attacks. This paper studied the results of cyber-attack interval forecasting with a pre-set intensity level of cyber-attacks. Interval forecasting is the forecasting of one interval from two predetermined ones in which a future value of the indicator will be obtained. For this, probability estimates of these events were used. For interval forecasting, a probabilistic neural network with a dynamic updating value of the smoothing parameter was used. A dividing bound of these intervals was determined by a calculation method based on statistical characteristics of the indicator. The number of cyber-attacks per hour that were received through a honeypot from March to September 2013 for the group 'zeppo-norcal' was selected as the indicator.