Network Intrusion Detection Using Supervised Machine Learning

Abstract

In the modern computer world, use of the
internet is increasing day by day. However, the
increasing use of the internet creates some security
issues. These days, such new type of security attacks
occurs every day and it is not easy to detect and prevent
those attacks effectively. One common method of attack
involves sending large amount of request to site or
server and server will be unable to handle such huge
requests and site will be offline for many days. This type
of attack is called distributed denial of service (DDOS)
attack, which act as a major security threat to internet
services and most critical attack for cyber security
world. Detection and prevention of Distributed Denial
of Service Attack (DDoS) becomes a crucial process for
the commercial organizations that uses the internet.
Different approaches have been adapted to process
traffic information collected by monitoring stations
(Routers and Servers) to distinguish malicious traffic
such as DDoS attack from normal traffic in Intrusion
Detection Systems (IDS). In general, Machine learning
techniques can be designed and implemented with the
intrusion systems to protect the organizations from
malicious traffic. Specifically, supervised clustering
techniques allow to effectively distinguishing the
normal traffic from malicious traffic with good
accuracy. In this paper, machine learning algorithms
are used to detect DDoS attacks collected from “KDD
cup 99 Dataset”, pre-processing and feature selection
technique is used on the dataset to enhance the
performance of the classifiers and reduce the detection
time. The classification algorithms such as C4.5
decision tree and Navie Bayes is applied on the training
dataset and the implementation of the algorithm is done
using spyder tool. The performance comparison of
algorithms is shown using confusion matrix and it is
found that C4.5 decision is more efficient in detection
of DDOS attack .The proposed method can be used as
DDoS defense system.