Detection and Characterization of DDoS Attacks using Time Based Features

Abstract

Abstract: In today's evolving cyber security environment, DDoS attacks are now one of the riskiest and most expensive attacks. DDoS detection and prevention has become critical as businesses and government agencies are affected by DDoS, which can disrupt network services and cause billions of dollars in damage. Our project offers a solution that detects different types of DDoS attacks using machine learning algorithms. The main aim of the project is to detect attacks of binary classification and multi class classification, this also reduces the time complexity. The proposed system studies the efficiency of time based features to detect and classify types of DDoS attacks using binary and multiclass classification. Using machine learning algorithms, the system analyzes this data to detect between legitimate and the attack network request. The addition of a time based features subset increases the accuracy of the system, reduces training time without compromising test accuracy. The general purpose of the project is not only detection research. It can be seen that training a subset of time-based features can reduce training time without affecting test accuracy; thus, the small subset of time-based features is itself useful for near-realtime applications with continuous learning.