Intrusion Detection System using Apache Spark Analytic System
Journal Title: IOSR Journals (IOSR Journal of Computer Engineering) - Year 2019, Vol 21, Issue 1
Abstract
In this study, an Intrusion Detection System (IDS) is proposed based on the use of machine learning and distributed computing. The proposed system uses classification techniques that are implemented in the built-in machine learning library in Apache Spark distributed computing framework. As the use of distributed computing allows the proposed method to provide rapid predictions for the packets flowing in the network, two classifiers are cascaded in order to combine their decisions for more accurate decisions. The Multi-Layer Perceptron (MLP) classifier is used as a binary classifier, where the output of this classifier only indicates whether the packet is a normal or attack packet. Packets predicted to be normal by this classifier are allowed through the network. However, packets predicted as attacks are classified again using the random forest classifier, which provides the state of the packet and the type of the attack as its output. If the packet is classified as a normal packet, it is also allowed to the network, otherwise it is filtered out. The results show that he proposed methodology has been able to improve the performance of IDS to 99.12%, which outperforms the state-of-the-art systems in the literature.
Authors and Affiliations
Asst. Prof. Dr. Sefer KURNAZ
Keywords
Related Articles
- EP ID EP441069
- DOI 10.9790/0661-2101013237.
- Views 125
- Downloads 0
Node Replication Attack Detection Algorithms in Wireless Sensor Networks: A Survey.
Abstract: The sensor networks are often deployed in hostile environments and are not attended for long time. Deployment of these networks is increased in the recent years, as they help in monitoring and analyzing d...
Assessing Buffering with Scheduling Schemes in a QoS Internet Router
A key requirement for service differentiation as required in Internet of the future and QoS to work effectively is the extension of traffic management routines of the current TCP/IP protocol. Two among of such &n...
Client Based System on Wireless Sensor Network for Efficient Packet Transmission
Abstract : Wireless sensor networks are vulnerable to the node distributor dividend, and several distributed protocols has been transfer the data in secure .Where MD5 algorithm that implements adaptive TTL, piggybacking,...
Interference Aware & SINR Estimation in Femtocell Networks
In wireless communication two main limitations are capacity and range. In the areas of high population density cellular service is far superior compared to scarcely populated areas. The initial cell...
Secure Development - Web Application Security.
In Current scenario, many Web applications go through rapid development phases like adopting agile methodology of development with very short turnaround time, making it difficult to identify and eliminate vulne...