Article Preview
Top1. Introduction
On demand service requirements in terms of software, platform, infrastructure, etc. has played a major role in the growth and evolution of the IT industry. There has been tremendous growth in this area since its inception based on the concept computing as Utility (Md Tanzim Khorshed, 2012). This to the IT world is cloud computing. With the growth of cloud computing, more and more players started providing services and the number of customers opting for services has grown exponentially. There are various reasons for companies adopting cloud computing services such as convenience in setup, on-demand capacity, requiring little maintenance and the most important of all highly dependable computing platforms (Naresh Kumar, 2012). With the growth in the adoption of cloud services, the security threats also further increased. The gaps in the area need to be minimized (Md Tanzim Khorshed, 2012).
This paper addresses the various security issues faced in a cloud environment and an attempt to predict the mitigation time to overcome different attacks are shown in Figure 1. It is very much essential to predict the mitigation time for different types of attacks since these attacks can unleash collateral damage to the network infrastructure and can disrupt the various services which in turn will disrupt the customer business. The data are provided by a leading Cloud Service provider based on a non-disclosure agreement and a total of seven different attack types are considered which is discussed in Section 1.1.
Figure 1. Overview of Cloud Computing (203)
1.1. Different Types of Attacks Considered in This Research
1.2. Contribution
This paper proposes the use of machine learning algorithms to predict the mitigation time from different types of attacks.
- ●
The statistical features of different attacks in a cloud based system are analyzed and the feature that directly impacts the mitigation time from the attack is identified.
- ●
Two different classes of machine learning algorithms are used in order to predict the mitigation time of the different types of attacks.
¡ Regression Based model
¡ Kernel Based model
2. Literature Review
(Zecheng He., 2017) et al. analyzed the existing strategies in tackling the denial of service (DOS) attacks. The existing passive defenses are not useful either in identifying the source of the attack or in acting based on attack statistical features. The authors proposed a DOS attack detection system where it uses machine learning algorithms to identify the attack in the cloud. The authors used statistical information on cloud servers as well as virtual machines and evaluated nine different machine learning algorithms to compare its performance. As per the analysis, more than 90 percent of the attacks under 4 different DOS attack categories are detected without degrading the performance (Zecheng He., 2017). The authors further stated that the statistical data can be used for a border analysis and detection of different attacks and prediction of different features.