Trojan Horse Infection Detection in Cloud Based Environment Using Machine Learning

Abstract

Cloud computing technology is known as a distributed computing network, which consists of a large number of servers connected via the internet. This technology involves many worthwhile resources, such as applications, services, and large database storage. Users have the ability to access cloud services and resources through web services. Cloud computing provides a considerable number of benefits, such as effective virtualized resources, cost efficiency, self-service access, flexibility, and scalability. However, many security issues are present in cloud computing environment. One of the most common security challenges in the cloud computing environment is the trojan horses. Trojan horses can disrupt cloud computing services and damage the resources, applications, or virtual machines in the cloud structure. Trojan horse attacks are dangerous, complicated and very difficult to be detected. In this research, eight machine learning classifiers for trojan horse detection in a cloud-based environment have been investigated. The accuracy of the cloud trojan horses detection rate has been investigated using dynamic analysis, Cukoo sandbox, and the Weka data mining tool. Based on the conducted experiments, the SMO and Multilayer Perceptron have been found to be the best classifiers for trojan horse detection in a cloud-based environment. Although SMO and Multilayer Perceptron have achieved the highest accuracy rate of 95.86%, Multilayer Perceptron has outperformed SMO in term of Receiver Operating Characteristic (ROC) area.

Web Attack Intrusion Detection System Using Machine Learning Techniques

“…The industry standard for calculating the error rate of a knowledge scheme on a specific dataset is the 10-fold cross-validation approach. For reliable results, tenfold cross-validation was used [40,42]. Moreover, the dataset is divided into ten portions (folds) for 10-fold cross-validation.…”

“…Preparing the features for machine learning is referred to as data preprocessing [39,40]. Data cleaning, feature selection, and data balancing are examples of data preparation procedures.…”

“…The datasets used in this study were divided into two halves for training and testing, using 10-fold validation. The 10-fold cross-validation method is commonly used to assess the error rate of a learning scheme on a specific dataset [40]. The dataset is divided into ten equal parts (folds) for 10-fold cross-validation, which assesses each part individually before averaging the results.…”

“…The dataset is divided into ten equal parts (folds) for 10-fold cross-validation, which assesses each part individually before averaging the results. Each data point in the dataset is utilized nine times for training and once for testing [40].…”