Development of hyper‐parameter‐tuned‐recurrent neural network for detection and mitigation of fraudulent resource consumption attack in cloud

Abstract

AbstractCloud computing is a paradigm that acts as an emerging technology with the intent of multimedia. Nevertheless, multimedia computing faces a financial burden. The essential characteristic of cloud computing is the cloud's pay‐as‐you‐go pricing model. Hence, several attacks are affecting cloud services since the threats are present long term. Such harmful attacks are named Fraudulent Resource Consumption (FRC) attacks. Due to this nature, attack detection is a critical task in cloud computing. To alleviate the problem, the proposed work is intended to detect and mitigate the FRC attack in the cloud without any vital issues. Initially, the HTTP web‐server logs are collected from the benchmark source, and the data preprocessing is performed for sequence generation, followed by the sequence decomposition process. Here, the Heuristic‐based Discrete Wavelet Transform is developed by the Adaptive Deer Hunting Optimization Algorithm (ADHOA) that is to be adopted for decomposing the sequence. Further, the hyper‐parameter Tuned‐Recurrent Neural Network (HT‐RNN) with estimating attack percentage is conducted for FRC detection. Once the detection of the FRC attack is done, then the mitigation procedure is processed for blocking them. Throughout the experimental analysis, the accuracy of the proposed ADHOA‐HT‐RNN method has attained 96%, and also, the precision of the proposed work offers 94%. The specificity of the designed model has secured 92%. The performance is assessed, and its result proves that the recommended system exploits the better detection performance. Thus, the simulation outcome has shown that the proposed ADHOA‐HT‐RNN model has attained superior performance compared to the other conventional approaches.