ENFORCING SENSITIVITY AND PRESERVATION IN DATA SECURITY AND PRIVACY USING MACHINE LEARNING

The exponential growth of sensitive data in digital ecosystems has necessitated robust security and privacy-preserving mechanisms to mitigate risks associated with unauthorized access, data breaches, and adversarial attacks. This study presents a machine learning-driven sensitive data security and privacy-preserving system that leverages federated learning, differential privacy, and secure multi-party computation to enhance protection while maintaining data utility. Traditional security frameworks often compromise either privacy or analytical utility, necessitating the development of a tunable system capable of balancing both dimensions. The proposed framework integrates XGBoost and Random Forest classifiers to enable adaptive privacy controls, reducing communication overhead and preserving model accuracy. The research adopts quantitative experimental methodologies, utilizing cryptographic techniques and privacy-aware machine learning models to safeguard sensitive data in healthcare, finance, and other critical domains. Comparative evaluations against existing systems reveal that the proposed model achieves higher privacy preservation while maintaining 98.29% accuracy for sensitive data classification. Furthermore, differential privacy accounting, federated analytics, and adversarial threat modeling are employed to assess real-world applicability. The study contributes to the advancement of context-aware security frameworks, offering scalable and customizable solutions for privacy-preserving data analytics. Future research will explore hybrid encryption mechanisms and real-world deployments to further enhance system robustness and applicability across diverse data environments.