Detection of DDoS Vulnerability in Cloud Computing Using the Perplexed Bayes Classifier

Table 1

Comparison of various DDoS attack studies.

Author	Year	Description	Remarks
Berguig et al. [20]	2018	1. The author of this work chose the KDD-CUP-99 data set.	The mobile-based strategies have been focused on resisting the DDoS attacks; however, the web-based strategies that were not covered could have also been covered.
		2. The authors provide the most extensively used mobile agent-based DDoS flooding assault defence tactics, a unique denial-of-service filter system based on mobile agents and naïve Bayes filters.
Nandi et al. [21]	2020	1. The authors of this work had chosen the essential characteristics from the NSL-KDD data set.	The authors did not attempt to create a DDoS detector with actual traffic in a real-world cloud system.
		2. The paper employed a hybrid technique in which a five-feature selection algorithm chooses and ranks the top most significant characteristics from the whole feature set.
Kim et al. [22]	2020	1. This study developed an intrusion model. Deep learning identifies DDoS attacks using the KDD-CUP 1999 data set and CSE-CIC-IDS 2018.	The data sets chosen for implementation also contain other classes of attacks. Hence, multiclass classification is not implemented in the current research.
		2. The implementation considered four attack types: DDoS, U2R, R2L, and probing.
		3. The machine learning technique CNN, which is further compared with RNN, has been used.
Cil et al. [23]	2021	1. This research uses deep neural network (DNN) to detect DDoS attacks on packet samples captured from network traffic.	It can create data sets like the CIC DDoS 2019 data set. It may be able to classify real-time DDoS attacks. By utilizing the data set, DNN and deep learning replicates will be built.
		2. The implementation is carried out with CIC DDoS 2019 data set to contain current DDoS attacks.
		2. Feature extraction, the classification process of the structure, is done to train the data set to the model.
Rangapur et al. [24]	2022	1. In this research, DDoS attacks are detected by using neural networks.	The data set consisting of different classes could be taken for implementation to improve the model’s efficacy.
		2. The main focus is to flag malicious and legitimate data flow and to prevent network performance degradation.
Saroha and Singh [25]	2019	1. The paper provides a qualitative analysis of all possible cloud vulnerabilities on each service model.	This study does not look at integrating into a cloud environment. No implementation was done for robust cloud systems. Also, the works do not use an ML algorithm.
		2. They have also proposed a countermeasure to enhance the security in cloud computing.
		3. Characterization of vulnerabilities has been presented.
Goel et al. [26]	2014	1. The author discussed cloud security vulnerabilities, dangers posed by a distributed denial-of-service (DDOS) assault on cloud computing infrastructure, and methods and tactics for detecting and preventing such attacks.	The paper had concentrated more on detection but not on mitigation.
		2. The author focused on and suggested an integrated and comprehensive model based on an intrusion detection system that addressed both internal misuse and external intrusion and that will detect or report the alert and vigorously challenge the attacks, reducing the overall risk of DDoS attacks.
Deshmukh et al. [27]	2015	1. The author discussed DDoS attacks, their impact on cloud computing, and the factors to consider when picking DDoS security systems.	VM attacks may degrade cloud performance, result in financial losses, and impact other servers in the same cloud architecture.
		2. The author gave a quick overview of DDoS assaults, followed by a taxonomy of attacks, kinds of attacks, and several countermeasures to reduce DDoS attacks.
Masdari and Jalali [28]	2016	1. The author has conducted an in-depth examination of the numerous forms of DDoS attacks suggested for the cloud computing environment, classifying them according to the cloud components or services they target.	There is no distinction between flash crowds and DoS assaults in clouds with dynamic material.
		2. It also included a thorough examination of the vulnerabilities used in various DoS assaults and an examination of the state-of-the-art solutions published in the literature for preventing, detecting, and dealing with each kind of DoS attacks in the Cloud.
Oberoi [29]	2017	1. The author investigated various security attacks (in general) concerning clouds.	This study does not offer a system to identify harmful insider assaults in cloud-based settings with accuracy and timeliness.
		2. Insider threat assaults should not be taken lightly, according to the available literature (research papers, reports, etc.).
		3. These assaults should not be taken lightly. The companies explicitly define the many categories of people capable of launching insider attacks and the dangers they face.
JeyaJothi et al. [30]	2022	1. In this study, to achieve higher quality classification, the fast correlation-based feature selection (FCBF) method was used for data preprocessing and further to remove irrelevant and redundant features of the data.	This has a limitation as it selects some limited features of the data set. The data pre-preprocessing could be done in a better way. Any new classifier may be used to achieve the best result.
		2. SVM classification has been done using a linear approach.
		3. Its limitation to dependent feature, which carries investigations, carried out feature extraction and its optimization techniques for OSA detection.