Fusion of SP-VAE and IMP-VAE for Proxy Attack Detection in ECommerce Systems

With the rapid development of e-commerce, proxy attacks, as a covert and efficient means of fraud, have seriously damaged fair competition and consumer trust in the market. Traditional detection methods often have low efficiency and high false positive rates, so dealing with complex and variable switching attacks requires tremendous effort. This article delves into the issue of proxy attack detection in e-commerce and proposes an innovative solution that integrates SP-VAE and IMP-VAE algorithms. By optimizing the network structure and introducing advanced mechanisms, IMP-VAE enhances the model's ability to handle highdimensional sparse data and improves the accuracy of feature extraction. Specifically, the model first uses IMP-VAE to extract deep features from e-commerce transaction data to capture hidden information that is crucial for detecting trust attacks. Then, the extracted features are further screened and compressed using SPVAE sparse constraints to remove redundant information and highlight anomalous features. The attack detection model combining SP-VAE and IMP-VAE provides a new method for security protection research in the field of e-commerce, which has important theoretical significance and practical application value. The experimental results show that the SP-VAE algorithm achieved a detection accuracy of 92.3% in detecting users supporting attacks, which is about 15 percentage points higher than traditional methods.