SIGNATURE BASED DETECTION FOR CLOUD BASED LOG MANAGEMENT USING MACHINE LEARNING TECHNIQUE

This paper presents signature based detection mechanism for cloud based log management using machine learning technique. The study aimed at detecting unauthorized log entry files into the cloud server and isolate from the network. To achieve this, data was collected from INFN-Tier data center during the hadron collider experiments and then used to train a neural network algorithm after processing using service-specific procedures. The performance was evaluated using accuracy and loss parameters and the result reported a training accuracy of 0.94188 and loss of 0.385 respectively. Finally after cross validation, the accuracy recorded was 0.915. The neural network was further compared with other state of the art algorithms such as Naïve Bayes, K-mean and Isolation forest. The Neural Network algorithm emerged as the most accurate among the tested algorithms, with an accuracy of 0.9155, indicating that it correctly predicted outcomes with an approximate success rate of 91.55%.