Class imbalance aware drift identification model for detecting diverse attack in streaming environment

Abstract

Detecting fraudulent transactions in a streaming environment presents several challenges including the large volume of data, the need for real-time detection, and the potential for data drift. To address these challenges a robust model is needed that utilizes machine learning techniques to classify transactions in real-time. Hence, this paper proposes a model for detecting fraudulent transactions in a streaming environment using xtream gradient boost (XGBoost), cross-validation and class imbalance aware drift identification (CIADI) model. The performance of the proposed method is evaluated using datasets named credit card and Network Security Laboratory (NSL-KDD) dataset. The results demonstrate that the model can effectively detect fraudulent transactions with high accuracy, recall, and F-measure. The results show that the proposed CIADI model attained 95.63% for the credit card dataset which is higher accuracy in comparison to the generative-adversarial networks (GAN), network-anomaly-detection scheme-based on feature-representation and data-augmentation (NADS-RA) and feature-aware XGBoost (FA-XGB). Further the proposed CIADI model attained 98.5% for the NSL-KDD dataset which is higher accuracy in comparison to the NADS-RA, stacked-nonsymmetric deep-autoencoder (sNDAE) and convolutional neural-network (CNN). This study suggests that the proposed method can be an effective model for detecting fraudulent transactions in streaming environments.