Hybrid Algorithm to Detect DDoS Attacks in VANETs

• Published in: Wireless personal communications Vol. 114; no. 4; pp. 3613 - 3634
• Main Authors: Adhikary, Kaushik, Bhushan, Shashi, Kumar, Sunil, Dutta, Kamlesh
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.10.2020; Springer Nature B.V
• Subjects: Algorithms; Communications Engineering; Computer Communication Networks; Denial of service attacks; Engineering; Kernels; Mobile ad hoc networks; Model testing; Networks; Security management; Signal,Image and Speech Processing; Vibration; Wireless networks; VANETs; SVM kernels; DDoS attacks; DDoS attack detection; Hybrid algorithm; Machine learning
• ISSN: 0929-6212; 1572-834X
• DOI: 10.1007/s11277-020-07549-y

Security and safety are fundamental issues in any wireless network. The problem becomes serious when the specified network is Vehicular Adhoc Network (VANET). VANET faces Distributed Denial of Service (DDoS) attacks, when several vehicles carry out various types of Denial of Service (DoS) attacks to disrupt the normal functioning of network, thereby endangering human lives. A highly efficient and reliable algorithm is required to be developed to detect and prevent DDoS attacks in VANET. This paper presents a hybrid detection algorithm based on the SVM kernel methods of AnovaDot and RBFDot for detecting DDoS attacks in VANETs. In this hybrid algorithm, features like collisions, packet drop, jitter etc. have been used to simulate real time network communication scenario where the network is operating under normal conditions, as well as under DDoS attacks. These features are used both for training and for testing the model based on the proposed hybrid algorithm. The performance of the model based on the proposed hybrid algorithm is compared with the models based on single SVM kernel algorithms AnovaDot and RBFDot based on Accuracy, Gini, KS, MER and H. The experimental results show that the model based on the proposed hybrid algorithm is superior to detect DDoS attacks as compared to the models based on single SVM kernel algorithms AnovaDot and RBFDot. The results also prove that by combining the the SVM kernel algorithms, an efficient and effective hybrid algorithm can be developed.