Chaotic physical security strategy based on manifold learning-assisted GANs for SDM-OFDM-PONs

Area of Science:

• Optical Communications
• Network Security
• Machine Learning

Background:

• Passive optical networks (PONs) are crucial for broadband access, but security remains a challenge.
• Spatial Division Multiplexing (SDM) and Orthogonal Frequency Division Multiplexing (OFDM) enhance capacity in PONs.
• Encryption is vital for securing high-speed data transmission in these networks.

Purpose of the Study:

• To propose a high-security encryption scheme for SDM-OFDM-PON systems.
• To leverage manifold learning-assisted generative adversarial networks (MFGANs) for robust encryption.
• To achieve a large key space and efficient encryption process.

Main Methods:

• Utilizing MFGANs to generate chaotic sequences for masking constellation and frequency.
• Applying manifold learning to capture complex structures from diverse chaotic models.
• Implementing parallel computing with Graphics Processing Units (GPUs) for accelerated encryption.

Main Results:

• Demonstrated a 70 Gb/s encrypted OFDM signal transmission over a 2 km, 7-core fiber.
• Achieved a key space of 1 × 10^183, significantly enhancing security.
• Reduced encryption time to approximately 1.38% of conventional methods using GPUs.

Conclusions:

• The proposed MFGAN-based encryption scheme offers high security and efficiency for SDM-OFDM-PONs.
• The method provides a large key space and significantly faster encryption times.
• This approach is suitable for next-generation high-speed and secure optical communication systems.