Privacy-preserving data analysis using a memristor chip with colocated authentication and processing.

Area of Science:

• * Neuromorphic Engineering
• * Applied Physics
• * Cybersecurity

Background:

• * Current privacy-preserving methods in healthcare face hardware and energy limitations in resource-constrained medical devices.
• * Existing solutions often separate security modules from analysis and memory from computation, leading to inefficiencies.
• * Sensitive patient data requires robust protection during monitoring and diagnostics.

Purpose of the Study:

• * To introduce a novel memristor-based system for integrated security and data analysis in healthcare.
• * To overcome the hardware and energy overheads of conventional privacy-preserving techniques.
• * To enhance privacy preservation and computational efficiency in medical devices.

Main Methods:

• * Development of a memristor-based colocated authentication and processing (CLAP) system.
• * Embedding physical unclonable functions within a compute-in-memory architecture.
• * Application of a differential stochastic mapping method based on information theory principles.

Main Results:

• * CLAP system demonstrated on a 130-nanometer memristor chip.
• * Achieved high device authentication accuracy (99.46% AUC) in an electrocardiogram data collection task.
• * Showcased significant energy efficiency gains (146.0-fold) and area reduction (17.6-fold).

Conclusions:

• * CLAP offers an intrinsically secure hardware solution for privacy-preserving data analysis in healthcare.
• * The system enhances both privacy preservation and computational efficiency for medical applications.
• * This approach is versatile and suitable for resource-limited medical devices.