Low Power Optoelectronic Neuromorphic Memristor for In-Sensor Computing and Multilevel Hardware Security...

Area of Science:

• Optoelectronics
• Neuromorphic Computing
• Cybersecurity

Background:

• Conventional encryption methods struggle with power efficiency and security vulnerabilities.
• Neuromorphic computing hardware offers a promising alternative for advanced encryption.
• Optoelectronic devices present unique opportunities for hardware-level security solutions.

Purpose of the Study:

• To develop a hardware-level multi-dimensional encryption paradigm using optoelectronic neuromorphic devices.
• To demonstrate a secure and energy-efficient encryption mechanism for digital information.
• To explore applications in motion detection, in-sensor computing, and secure communication.

Main Methods:

• Encoding ASCII characters into optical pulse sequences (wavelength, duration, pulse number).
• Utilizing optoelectronic neuromorphic devices for physically obfuscated electrical responses.
• Training a convolutional neural network for signal decryption and authentication.
• Constructing a dual-authentication system for maritime communication scenarios.

Main Results:

• Achieved low energy consumption (3.3 fJ) with the optoelectronic neuromorphic system.
• Demonstrated high recognition accuracy (97.4%) for legitimate users during decryption.
• Showcased robustness against unauthorized access with low accuracy (∼2.88%).
• Successfully implemented a two-layer authentication system for maritime communication.

Conclusions:

• The developed neuromorphic optoelectronic system provides a secure and energy-efficient encryption solution.
• The system shows significant potential for next-generation neuromorphic maritime communication systems.
• Integration of motion perception, reservoir computing, and photonic encryption enhances communication security.