Abstract

Internet and mobile computing have significantly improved performance in various applications, including digital payments, storage, and confidential information access. However, 44% of frauds were identified in various fields from 1998 to 2022, making security and confidentiality crucial. To address this issue, researchers have developed and implemented cloud-based security systems using deep and machine learning optimization methods. These systems achieve high performance, making them a major requirement in cloud computing design. For credit card fraud identification, fully convolution neural networks (FCCN) and Dynamic Multi-Object based Optimization for Gradient Boosting algorithm (DMOGB) technologies are implemented. These technologies enable real-time fraud identification, classification, and feature extraction. The approach operates on clouds, with a central decision and privacy-preserving mechanism, making fraud identification easier. To decrease the fraud ratio, a real and accurate fraud detection system is needed. This research uses deep and machine learning optimization methods to detect credit card fraud. Existing works have limited accuracy, F-score, recall, and precession. To address these limitations, the research introduces deep learning mechanisms like fully convolution neural networks, convolution neural networks, and machine learning mechanisms like SVM, neural network, and LR methods. The experiment is performed on Amazon Web Services Cloud, and performance measures such as accuracy, recall, and true positive rate are calculated. The implementation results in improvements in accuracy, decision rate, and false alarm rate.