Big data security in radiation and healthcare analytics: Challenges and solutions

Healthcare data related to radiation are particularly high-value data, containing sensitive information related to the clinical, institutional and operational aspects of these data. Based on technical, organizational and regulatory predictors, this study designed and tested machine learning models that predict and classify Radiation Data Security into three levels of security. The variables that were considered were: Encryption Strength, Access Control Mechanisms, Data Anonymization, System Architecture Complexity, Machine Learning Threat Detection, Regulatory Compliance, Data Integrity and Quality, Security Protocols, Data Sharing Mechanisms, Auditing and Logging Systems, Data Backup and Recovery Systems, Cloud Service Providers’ Security, User Awareness and Training, and Real-Time Threat Monitoring. The supervised models used were Decision Tree, Support Vector Machine and Random Forest. Confusion matrix, accuracy, Kappa statistics, precision, recall, F1-score, specificity, and balanced accuracy were used to measure performance. The highest accuracy was observed by Decision Tree (97.31%) followed by Random Forest (96.85%) with SVM having lower performance especially in class 2. The results suggest good internal classification accuracy for the current data set, but further external validation is needed to confirm generalizability to other healthcare settings.