Data security system design using crypto-steganography in a network-based application

Zero-Day attacks take advantage of vulnerabilities that are not known, thus making it difficult to identify them with standard methods. As a result of this, organizations are unable to minimize the potential risks of these attacks which often lead to financial and reputational damage to the organization. In order to enhance zero day threat detection, this study proposes a hybrid anomaly detection model that combines Isolation Forest and XGBoost machine learning techniques. The model was trained on 80% of the TII-SRC-23 dataset and tested on 20% of the overall dataset. The hybrid model achieved an accuracy of 99.95%, precision of 99.91%, recall of 99.99% and F1-Score of 99.95%. These results highlight the effectiveness of combining anomaly-based detection with ensemble classification. Compared to the standard detection techniques, this combination offers a significant improvement. Key performance metrics including precision, recall, F1-score, and False Alarm Rate were used for evaluation, demonstrating the model's capacity to reduce false alarms while ensuring accurate and timely identification of zero-day attacks.