A learning-based hybrid framework for detection and defence of DDoS attacks. (2017)

• Record Type: Journal Article
• Authors: Subbulakshmi, T.
• Is Part Of: International journal of internet protocol technology. Volume 10:Number 1 (2017)
• Page Start: 51
• Page End: 60
• Publication Date: 2017
• Subjects: DDoS attacks -- attack source identification -- IP addresses -- entropy based defence -- back propagation neural networks -- BPNN -- self-organising maps -- SOM -- enhanced SVM -- support vector machines -- ESVM -- machine learning -- distributed DoS -- denial of service -- cyberattacks; File Transfer Protocol (Computer network protocol) -- Periodicals; Multicasting (Computer networks) -- Periodicals; 004.678
• Languages: English
• ISSNs: 1743-8209

Abstract

Distributed denial of service (DDoS) attacks are those which deplete the valuable resource available for the legitimate user and reduces the business value of any web service provided. This sort of cyber-attacks has to be detected and respective actions have to be taken on them. An integrated detection and defensive mechanism is proposed in this paper to generate and detect DDoS attacks using machine learning algorithms such as back propagation neural network (BPNN), self-organising map (SOM) and enhanced support vector machine (ESVM) and to identify the real IP address of the spoofed attack source using the entropy-based defensive mechanism. The detection and defence mechanism are found to be effective in identifying the attack source with 99% accuracy using ESVM and response time of less than two seconds using the entropy-based tracing scheme. The real source of attacks is filtered using the IP tables to defend the DDoS attacks.