DDoS Detection System: Using a Set of Classification Algorithms Controlled by Fuzzy Logic System in Apache Spark

被引:50
作者
Alsirhani, Amjad [1 ]
Sampalli, Srinivas [1 ]
Bodorik, Peter [1 ]
机构
[1] Dalhousie Univ, Fac Comp Sci, Halifax, NS B3H 1W5, Canada
来源
IEEE TRANSACTIONS ON NETWORK AND SERVICE MANAGEMENT | 2019年 / 16卷 / 03期
关键词
DDoS attack; DDoS detection; fuzzy logic system; machine learning; classification algorithms; apache spark; apache Hadoop; cloud computing; TREES;
D O I
10.1109/TNSM.2019.2929425
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Distributed denial of service (DDoS) attacks are a major security threat against the availability of conventional or cloud computing resources. Numerous DDoS attacks, which have been launched against various organizations in the last decade, have had a direct impact on both vendors and users. Many researchers have attempted to tackle the security threat of DDoS attacks by combining classification algorithms with distributed computing. However, their solutions are static in terms of the classification algorithms used. In fact, current DDoS attacks have become so dynamic and sophisticated that they are able to pass the detection system thereby making it difficult for static solutions to detect. In this paper, we propose a dynamic DDoS attack detection system based on three main components: 1) classification algorithms; 2) a distributed system; and 3) a fuzzy logic system. Our framework uses fuzzy logic to dynamically select an algorithm from a set of prepared classification algorithms that detect different DDoS patterns. Out of the many candidate classification algorithms, we use Naive Bayes, Decision Tree (Entropy), Decision Tree (Gini), and Random Forest as candidate algorithms. We have evaluated the performance of classification algorithms and their delays and validated the fuzzy logic system. We have also evaluated the effectiveness of the distributed system and its impact on the classification algorithms delay. The results show that there is a trade-off between the utilized classification algorithms' accuracies and their delays. We observe that the fuzzy logic system can effectively select the right classification algorithm based on the traffic status.
引用
收藏
页码:936 / 949
页数:14
相关论文
共 40 条
  • [1] Trees vs Neurons: Comparison between random forest and ANN for high-resolution prediction of building energy consumption
    Ahmad, Muhammad Waseem
    Mourshed, Monjur
    Rezgui, Yacine
    [J]. ENERGY AND BUILDINGS, 2017, 147 : 77 - 89
  • [2] Dynamic Fuzzy-Logic Based Path Planning for Mobility-Assisted Localization in Wireless Sensor Networks
    Alomari, Abdullah
    Phillips, William
    Aslam, Nauman
    Comeau, Frank
    [J]. SENSORS, 2017, 17 (08)
  • [3] [Anonymous], 2018, P 2018 9 IFIP INT C
  • [4] [Anonymous], 2013, AFRICAN J COMPUT ICT
  • [5] [Anonymous], P INT C CLOUD COMPUT
  • [6] [Anonymous], CYBERNETICS INFORMAT
  • [7] [Anonymous], PATTERN RECOGN LETT
  • [8] [Anonymous], P IEEE IFIP NOMS IEE
  • [9] Balkanli E, 2015, 2015 IEEE 40TH LOCAL COMPUTER NETWORKS CONFERENCE WORKSHOPS (LCN WORKSHOPS), P611, DOI 10.1109/LCNW.2015.7365905
  • [10] Random forests
    Breiman, L
    [J]. MACHINE LEARNING, 2001, 45 (01) : 5 - 32