Abstract: The increasing adoption of cloud-based healthcare systems has introduced significant cybersecurity challenges, necessitated robust threat detection mechanisms while preserved patient data privacy. Traditional security solutions often struggle to detect sophisticated cyber threats and protect sensitive healthcare data. Cloud-based healthcare systems face increasing cybersecurity threats, including unauthorized access, data breaches, and malicious intrusions. Traditional AI-based threat detection models struggle with imbalanced data and lack robust privacy mechanisms, risking sensitive patient information exposure. Existing security solutions often fail to detect sophisticated cyber threats while maintaining privacy. There is a critical need for an effective, scalable, and privacy-preserving cybersecurity threat detection model. This research proposes a privacy-preserving cybersecurity threat detection model using Long Short-Term Memory (LSTM) networks optimized with Differentially Private Stochastic Gradient Descent (DP-SGD). The LSTM model effectively identifies anomalies in security logs, authentication records, and network traffic, while DP-SGD ensures privacy by introducing controlled noise during training. The proposed approach enhances accuracy, security, and scalability in cloud-based healthcare environments. Experimental results demonstrate high performance, achieving an accuracy of 94%, precision of 91%, recall of 89%, and F1-score of 90%. Additionally, an AUC-ROC score of 1.00 confirms its strong classification capability. The model efficiently scales with increasing data volume, ensuring real-time threat detection and adaptive security measures. This study provides a scalable, privacy-preserving, and effective solution for mitigating cybersecurity threats in cloud healthcare systems.

Keywords: Cloud-based healthcare, cybersecurity, threat detection, Long Short-Term Memory (LSTM), Differentially Private Stochastic Gradient Descent (DP-SGD), anomaly detection, privacy preservation, deep learning, security logs, network traffic analysis, scalability.

| DOI: 10.17148/IJIREEICE.2020.8713