In recent years the usage of internet and cloud has become essential by generating substantial information sharing over the network. This scenario has developed an opportunity to the attackers to perform malicious activities on the network to gain unauthorised access on the data stored inside the server. As per the existing security policies the experts are using various security measures that includes physical security to the system, Firewalls and utilization of trusted data base software. But all these security methods are mostly being compromised by attackers. Hence to provide security dynamically by understanding the type of malicious attacks many researchers applied machine learning techniques. Machine learning techniques are used to decide whether user connection request is malicious or normal by analysing features of the network connection. In this paper we are introducing a machine learning model that select the best features required to classify the connection. In addition, comparative analysis is performed among various classification algorithms to find suitable in identifying attacks. In the model NSL-KDD data set is used which consists of 41 features and more than 50,000 records which are labelled across four types of attacks such as DOS, PROBE, R2L, U2R. This Data Set is divided into training and testing and trained to Decision Tree Classifier. The accuracy of the model is compared with existing techniques. This Intruder Detection System found effective to detect malicious connections.
Volume 11 | Issue 11
Pages: 221-228
DOI: 10.5373/JARDCS/V11I11/20193191