Stochastic switching and analog-state programmable memristor and its utilization for homomorphic encryption hardware

Area of Science:

• Materials Science
• Computer Engineering
• Cryptography

Background:

• Homomorphic encryption (HE) enables secure computation on encrypted data, crucial for cloud-edge communication.
• Edge devices require low-power, compact homomorphic encryption hardware (HE-HW).
• Existing HE-HW solutions face challenges in power consumption and form factor.

Purpose of the Study:

• To propose a novel memristor device for efficient HE-HW.
• To demonstrate the feasibility of using memristors for encryption key generation and data processing.
• To develop a prototype HE-HW using a 1-transistor-1-memristor (1T1M) array.

Main Methods:

• Fabrication of a Pt/Ta2O5/Mo metallic cluster-type memristor (Mo-MCM).
• Characterization of the Mo-MCM's stochastic set-switching behavior for random number generation.
• Utilizing the device's analog non-volatile memory characteristics for data storage.
• Integration of Mo-MCMs into a 1T1M array for prototype HE-HW construction.

Main Results:

• The Mo-MCM exhibits inherent stochastic set-switching, suitable for encryption key generation.
• The device accurately stores analog conductance states, functioning as a non-volatile analog memristor.
• Simultaneous encryption key generation, data encryption, and decryption were achieved within a single device.
• A prototype HE-HW was successfully demonstrated using the Mo-MCM based 1T1M array.

Conclusions:

• The proposed Mo-MCM offers a promising solution for low-power, compact HE-HW.
• In-memory computing with memristors enables efficient and integrated encryption functionalities.
• This approach paves the way for secure edge computing applications.