A Review on the Optimization of Enterprise Data Security Using Neural Networks

Resumen

Technology in business field has helped solve problems that involve card fraud or commercial transactions; however, it has also generated a great need to guarantee the security of customer data. The use of neural networks has emerged as an indispensable tool to address this type of problem. In this study, a systematic review was developed of 64 articles from Scopus and Web of Science, to identify the type of business data, the neural network architecture used, and the influence on a type of analyzed business problem. Search strategies such as PICO and PRISMA2020 were combined for the selection. Based on inclusion and exclusion criteria, this paper seeks to answer the research question: How the neural network models optimize enterprise data security? Four metrics were identified: Accuracy, Precision, Recall, and F-Score, to compare the efficiency of neural network models such as CNN, DNN, LSTM, or their variants. These models outperformed other current techniques in the identification of card or transaction fraud; however, their implementation remains limited in small and medium-sized enterprises due to restrictions in technological infrastructure and computational capacity.