E-ComShield: Enhancing E-Commerce Cybersecurity through Cloud-Native Threat Detection Frameworks

Abstract

With the surge of fast-growing digital business, threats to e-commerce websites have grown in terms of 
complexity and volume. This paper proposes E-ComShield, a new cloud-native cybersecurity solution that is 
specifically developed to actively monitor, categorize, and counter advanced threats within e-commerce 
systems. With an additional implementation using an Adaptive Hybrid Deep Learning model integrated with 
Federated Feature Fusion (AHDL-FFF), E-ComShield combines Convolutional Neural Networks (CNN), Long 
Short-Term Memory networks (LSTM), and Deep Neural Networks (DNN) to learn spatial, temporal, and 
abstract patterns in the heterogeneous transactional data. The system utilizes sophisticated data preprocessing 
methods, such as missing value imputation, outlier detection based on Z-scores, label encoding, and Min-Max 
scaling, to improve model training and improve detection accuracy. Federated learning maintains privacy
preserving collaboration among distributed data nodes, ensuring user confidentiality without degrading 
analytical 
performance. 
Deep experiments prove that E-ComShield outperforms, registering 99.65% accuracy, precision, recall, and F1
score superior to common benchmarks like SVM, Naive Bayes, and isolated CNN. Additionally, heat map 
examination of threat category types and ranking importance of features reinforces the model in distinguishing 
among variable attack vectors such as DDoS, session hijacking, and bot outliers with limited misclassification. 
Cloud-native deployment through CI/CD pipelines ensures scalability, real-time self-adaptation, and dynamic 
threat intelligence integration. Results corroborate E-ComShield as an end-of-the-art protection solution against 
current e-commerce websites' dynamic security threats. Future research direction of autonomous, self-healing, 
and federated defence platforms is opened in this work for the digital commerce environments.