Detecting DDoS Attacks Using Polyscale Analysis and Deep Learning

Distributed﻿denial-of-service﻿(DDoS)﻿attacks﻿are﻿serious﻿threats﻿to﻿the﻿availability﻿of﻿a﻿smart﻿grid﻿ infrastructure﻿services﻿because﻿they﻿can﻿cause﻿massive﻿blackouts.﻿This﻿study﻿describes﻿an﻿anomaly﻿ detection﻿method﻿for﻿improving﻿the﻿detection﻿rate﻿of﻿a﻿DDoS﻿attack﻿in﻿a﻿smart﻿grid.﻿This﻿improvement﻿ was﻿achieved﻿by﻿increasing﻿the﻿classification﻿of﻿the﻿training﻿and﻿testing﻿phases﻿in﻿a﻿convolutional﻿neural﻿network﻿(CNN).﻿A﻿full﻿version﻿of﻿the﻿variance﻿fractal﻿dimension﻿trajectory﻿(VFDTv2)﻿was﻿ used﻿to﻿extract﻿inherent﻿features﻿from﻿the﻿stochastic﻿fractal﻿input﻿data.﻿A﻿discrete﻿wavelet﻿transform﻿ (DWT)﻿was﻿applied﻿to﻿the﻿input﻿data﻿and﻿the﻿VFDTv2﻿to﻿extract﻿significant﻿distinguishing﻿features﻿ during﻿data﻿pre-processing.﻿A﻿support﻿vector﻿machine﻿(SVM)﻿was﻿used﻿for﻿data﻿post-processing.﻿The﻿ implementation﻿detected﻿the﻿DDoS﻿attack﻿with﻿87.35%﻿accuracy.


ReSULTS AND DISCUSSIoN
In this section, the justification for each step and the results of the proposed anomaly detection algorithmareexplained.
Whilemostoftheworkhasbeendoneonimages,one-dimensionalsignals(e.g.,stockpriceindex) arebeingnowconsideredtoo(e.g., Lou,Qi,&Li,J.,2018)(IEEE), a member of the Association of Computing Machinery (ACM), the American Association for the Advancement of Science (AAAS), the Engineers Geoscientists Manitoba (APEGM), and a member of other societies.For over 45 years, he has been very active all the IEEE levels: Region 7 (IEEE Canada), Council, Section, Chapter, and Student Branch. He was elected IEEE Canada President Elect 2014-2015, IEEE Director Elect (Region 7) 2014-2015, IEEE Canada President/Director 2016-2017, Past President 2018-19, and IEEE Educational Activities Vice President, 2018-19.He has been involved in research on cognitive systems, computational intelligence, robust real-time computing engines, and computer memories.Applications included biomedical, industrial monitoring and controls, aerospace, and space.He has authored and co-authored over 780 publications in the above areas, as well as supervised 73 Master's and Doctorate graduate students, over 200 undergraduate final-year thesis/ capstone project students, and mentored 35 summer research students.

Figure 1 .
Figure 1.The number of packets within a stationary frame size

Figure
Figure 2. The algorithm of DDoS attack detection

Figure 3 .
Figure 3.A single-stage convolutional neural network architecture

Figure 4 .
Figure 4. Optimal hyperplane to classify the positive and negative data with the support vector machine algorithm (After Ferens (2016))

Figure 5 .
Figure 5. Relationship between parameters n i , δ n , N T

Figure 6 .
Figure 6.A graphic illustration describing the concept of covering all points of data to calculate the VFDv2

Table 1 .
Figure 8.The trajectory regarding VFDv2 algorithm of the ATS data within the stationary frame and k low = 0

Figure 9 .
Figure 9.The trajectory regarding VFDv2 algorithm of the ATS data within the stationary frame and k low = 1 offeredananomalydetectionbasedonphaseangle,currentandvoltageinaframe.

Table 2 . Simulation result of the one-stage CNN as a processing step (Ghanbari et al., 2017) Table 3. Simulation result of the pre-processed data by the db4 and one stage CNN (Ghanbari et al., 2017) Table
Maryam Ghanbari is a PhD candidate in the Department of Electrical and Computer Engineering at the University of Manitoba, working with her advisor Professor Witold Kinsner.She obtained her Master's degree in Computer Software Engineering and her Bachelor's degree in Computer Software Engineering both from the Islamic Azad University, Qazvin Branch, Iran.Her current research focuses on the detection of vulnerabilities in smart grids that are under cyber domain and physical domain attacks, using polyscale analysis and convolutional neural networks with deep learning.Witold Kinsner is Professor in the Department of Electrical and Computer Engineering, University of Manitoba, Winnipeg, Canada.He obtained his Ph.D. degree in Electrical and Computer Engineering from McMaster University in 1974, and became Assistant Professor at McMaster University and then at McGill University.He was a co-founder and Director of Research of the Industrial Applications of Microelectronics Centre from 1979 to 1987.He is a Fellow of the Engineering Institute of Canada (FEIC), a Fellow of Engineers Canada (FEC), a Fellow of the Canadian Academy of Engineering (CAE), a Life Senior Member of the Institute of Electrical & Electronics Engineers