Applying Big Data Analytics in DDos Forensics: Challenges and Opportunities

DDoS (Distributed Denial-of-Service) attacks greatly affect the internet users, but mostly it’s a catastrophe for the organization in terms of business productivity and financial cost. During the DDoS attack, the network log file rapidly increases and using forensics traditional framework make it almost impossible for DDoS forensics investigation to succeed. This paper mainly focuses on finding the most suitable techniques, tools, and frameworks in big data analytics that help forensics investigation to successfully identify DDoS attacks. This paper reviewed numbers of previous research that related to the topic to find and understand general terms, challenges and opportunities of using big data in forensics investigation. The data mining tools used in this paper for simulation was RapidMiner because of its ability to prepare the data before the analysis and optimizes it for quicker subsequent processing, and the dataset used was taken from University of New Brunswick’s website. Algorithms that were used to evaluate the DDoS attack training dataset are Naïve Bayes, Decision Tree, Gradient Boost and Random Forest. The evaluation results projected that the majority of algorithms has above 90% of accuracy, precision and recall respectively. Using the data mining tools and recommended algorithms will help reduce processing time associated with data analysis, reduce cost and improve the quality of information. Future research is recommended to install in an actual network environment for different DDoS detection models and compare the efficiency and accuracy in real attacks.