Authors - Franck W. Boubayi Abstract - The rapid development of e-commerce platforms has significantly increased the exposure of digital systems to advanced cyber threats such as phishing, DDoS attacks, identity theft, and data breaches. To address the limitations of conventional security mechanisms, this paper proposes a Risk-Adaptive Multi-layer Security Framework (RAMSF) integrating multi-factor authentication, AI-based intrusion detection, post-quantum cryptography, and blockchain-based distributed auditing. The main contribution of the model lies in an adaptive decision engine based on dynamic risk evaluation, enabling real-time adjustment of security policies according to behavioral context and detected anomalies. The framework is evaluated using the CICIDS2017 and UNSW-NB15 datasets with 10-fold cross-validation. Experimental results show that RAMSF outperforms several classical models, including SVM, Random Forest, CNN, and XGBoost, achieving 97% accuracy, 95.5% F1-score, 98% AUC, and a low false positive rate. These results demonstrate that adaptive hybrid architectures represent a promising approach for strengthening cybersecurity in e-commerce systems against both current and post-quantum threats.
Open Zoom