A Novel Optimized Perturbation-Based Machine Learning for Preserving Privacy in Medical Data

Area of Science:

• Medical Informatics
• Machine Learning
• Cybersecurity

Background:

• Medical datasets require robust privacy protection due to sensitive patient information.
• Existing machine learning models often struggle to provide adequate data privacy in healthcare applications.
• Securing digital patient files in hospitals is a critical challenge.

Purpose of the Study:

• To develop a novel model for enhancing medical data privacy.
• To validate the performance of the proposed model in disease classification tasks.
• To incorporate specific modules for data security and privacy.

Main Methods:

• A Honey pot-based Modular Neural System (HbMNS) was designed and implemented.
• The HbMNS model integrates a perturbation function and a verification module for data privacy.
• Performance was evaluated using disease classification, with outcomes assessed before and after perturbation.
• A Denial of Service (DoS) attack was simulated to test system resilience.

Main Results:

• The HbMNS model demonstrated improved performance in disease classification with enhanced data privacy.
• The perturbation function and verification module effectively contributed to data security.
• The system showed resilience against simulated DoS attacks.
• Comparative analysis confirmed superior outcomes compared to existing models.

Conclusions:

• The proposed Honey pot-based Modular Neural System (HbMNS) offers a promising solution for medical data privacy.
• The integration of perturbation and verification modules enhances the security of medical datasets.
• HbMNS provides a more effective approach to data privacy in medical applications compared to other methods.