Privacy-Enhanced Algorithms for E-commerce Fraud Detection

This paper aims to improve the accuracy and privacy protection of credit card fraud detection within the E-commerce domain through differential privacy algorithms. First, we introduce the particular dataset under study with its preprocessing steps. Then in the model construction, we use four machine learning algorithms, logistic regression, random forest, XGBoost and AdaBoost, combined with differential privacy techniques and compare the accuracy of each algorithm in both original and differential privacy processing cases. The results show that the trade-off relationship between the accuracy rate of the algorithms with differential privacy processing and the stability of the model is the core of privacy protection under different parameters and complexity conditions. Finally, it is concluded that differential privacy is important in improving E-commerce security and privacy protection, and future research can be explored in more complex models, large-scale datasets, and more diverse domains to further optimize privacy protection strategies.