To Develop Machine Learning Model For Attack Detection And Analysis in Cloud

Abstract

Cyber assaults are a serious hazard to people and organizations in the modern digital environment. To stop data breaches and other cyber security issues, it is crucial to identify and analyses these assaults. In this research, a unique machine learning-based technique to attack analysis and detection is proposed. The suggested method analyses network, memory, and disk data using machine learning techniques to find patterns of behaviour that point to an assault. The usefulness of the suggested technique is assessed using real-world datasets, and the findings show that it is very accurate in identifying and analyzing assaults. The suggested method helps strengthen an organization's security posture and stop data breaches, which can cause severe financial and reputational harm. Moreover, the paper suggests a machine-learning approach for identifying and examining possible network, memory, and disk factors related to attack detection in the cloud. The model profiles virtual machine data due to hypervisor-level actions, enabling cloud-based detection and retention of a potential network, memory, and disk characteristics. The study uses the VM Resource Utilization Slope dataset, which includes metadata, network, memory, and disk components, to pinpoint the causes of illegal activity on a virtual machine. After analyzing the normal and attack data patterns for VM 5 and 6 dataset characteristics, we used the RF, XGBOOST, and NN algorithms to find relevant features. We then ran a binary classification task to detect illegal behavior, testing several machine learning models such as Decision Tree(DT), Boosting Decision Tree(Boosting DT), Bagging Decision Tree(Bagging DT), Gradient Boosting Classifier(GBC), Support Vector Machine(SVM), Random Forest(RF), k-nearest neighbors(KNN), Logistic Regression(LR), Adaptive Boosting(ADABOOST), extreme Gradient Boosting(XGBOOST), and MLP Classifier(MLPC). These models' performance was evaluated using two sets of data, unbalanced and balanced, with all models earning train, test and cross-validation accuracy scores ranging from 0.99 to 1.00. The proposed approach enhances the security and privacy of cloud-based virtual machine data gathering, facilitating more effective attack detection and analysis. Overall, this research proposes a promising approach to address the increasing threat of cyber-attacks in the digital age.