cqu_8554+ATTACHMENT04+ATTACHMENT04.3.pdf (2.01 MB)
Download file

A survey on gaps, threat remediation challenges and some thoughts for proactive attack detection in cloud computing

Download (2.01 MB)
The long-term potential benefits through reduction of cost of services and improvement of business outcomes make Cloud Computing an attractive proposition these days. To make it more marketable in the wider IT user community one needs to address a variety of information security risks. In this paper, we present an extensive review on cloud computing with the main focus on gaps and security concerns. We identify the top security threats and their existing solutions. We also investigate the challenges/obstacles in implementing threat remediation. To address these issues, we propose a proactive threat detection model by adopting three main goals: (i) detect an attack when it happens, (ii) alert related parties (system admin, data owner) about the attack type and take combating action, and (iii) generate information on the type of attack by analyzing the pattern (even if the cloud provider attempts subreption). To emphasize the importance of monitoring cyber attacks we provide a brief overview of existing literature on cloud computing security. Then we generate some real cyber attacks that can be detected from performance data in a hypervisor and its guest operating systems. We employ modern machine learning techniques as the core of our model and accumulate a large database by considering the top threats. A variety of model performance measurement tools are applied to verify the model attack prediction capability. We observed that the Support Vector Machine technique from statistical machine learning theory is able to identify the top attacks with an accuracy of 97.13%. We have detected the activities using performance data (CPU, disk, network and memory performance) from the hypervisor and its guest operating systems, which can be generated by any cloud customer using built-in or third party software. Thus, one does not have to depend on cloud providers' security logs and data. We believe our line of thoughts comprising a series of experiments will give researchers, cloud providers and their customers a useful guide to proactively protect themselves from known or even unknown security issues that follow the same patterns.

History

Volume

28

Issue

6

Start Page

833

End Page

851

Number of Pages

19

ISSN

0167-739X

Location

London, UK

Publisher

Elsevier B.V.

Language

en-aus

Peer Reviewed

Yes

Open Access

No

External Author Affiliations

Centre for Intelligent and Networked Systems (CINS); Faculty of Arts, Business, Informatics and Education;

Era Eligible

Yes

Journal

Future generation computer systems.