Anomaly Detection System in Cloud Computing Network

Author Name(s): S.Bharath Reddy, D.Malathi, S.Shijoe Jose
Author Email: bharath.bittu945@gmail.com

Abstract

Security in cloud computing poses higher challenges due to modern day distributed computing architectures. Cloud computing is a quickly developing IT show for the trade and conveyance of various administrations through the Internet. Be that as it may, there are a plenty of security worries in cloud computing which still should be handled (e.g., confidentiality, auditability and Privileged User Access). To recognize and avoid such issues, the Intrusion Detection System (IDS) and Intrusion Prevention System (IPS) are powerful instruments against assaults, for example, SQL Injection. Monitoring such a highly complex network environment needs a system capable of handling passive monitoring efficiently with less human intervene. Intrusion detection is a versatile security paradigm which can avert most of the computer and network related attacks if efficiently employed.  In this paper, we propose a modern security monitoring framework capable of detecting security attacks through data intelligence. The results of experiments prove that the average time taken to predict attacks by K-means algorithm is better than Naive Bayes, MLP, and SVM algorithms.

Introduction

Cloud computing has become ubiquitous computing platform for on-demand and cost-effective infrastructures. The exceptional effectiveness of the platform arises from the technology to share the underlying physical resources not limited to CPU Cores, RAM, Storage and Networking between users at the same time without compromising the performance. With the advent of emerging virtual machine technologies, cloud computing has a stronger growth in broad adoption. Virtualization technologies play a key role in fuelling the rapid growth of cloud computing platform. Virtualization allows resources on the single physical machine to be abstracted by a virtual layer enabling to group multiple virtual resources on different physical machine hardware’s as one massive virtual resource. Physical devices in a cloud computing infrastructure are connected with high speed wired links. When a new virtual machine is created with some specific resource requirement, the allocated virtual resource could be from different physical machine based on the virtual machine placement algorithms1.

Deployment strategies and different virtualization technologies possess various limitations and also open new areas of challenges for infrastructure reliability. Operation of cloud computing infrastructure mainly relays on networking of multiple physical resources. Due to the distributed computing resources, it is evident that a vulnerable virtual machine could significantly affect the overall cloud computing infrastructure performance. Network security for cloud computing environment possesses a significant degree of the challenge due to distributed nature of the infrastructure. In this paper, we explore the vulnerabilities that degrade the performance of cloud computing platform concerning the underlying networking infrastructure and proposed a framework for network security deployment using K-Means clustering algorithm.

Conclusion

In this work, we analyzed the techniques for anomaly detection in literature and proposed the solution for improve the performance of anomaly detection for modern day attacks. Finally, evaluated the proposed solution in a real time testbed. The datasets was collected from Canadian university to train the algorithm. K-means were able to get a best accuracy of 0.534 and got an average 0.00886 sec when used in real time environment. The system can be implemented on a variety of platforms, has a very simple approach, is easy to configure, does not incur any cost for implementation (as both routers and computer are already there in the network) and does not need any specialized person for its operation.

282 total views, 2 views today

Download PDF File

About the author: admin