Channel characteristics estimation based on a secure optical transmission system with deep neural networks

Area of Science:

• Cybersecurity
• Optical Communications
• Machine Learning

Background:

• Existing secure optical transmission systems face challenges like separate plaintext/key transmission and vulnerability to eavesdropping based on fiber length.
• Low key generation rates (KGR) and insecurity against eavesdroppers with local fiber length information are significant drawbacks.

Purpose of the Study:

• To propose a novel secure optical transmission system utilizing neural networks (NNs) for enhanced security and efficiency.
• To overcome limitations of existing systems by enabling simultaneous plaintext and key transmission and dynamic key transformation.

Main Methods:

• Employing neural networks (NNs) to estimate channel characteristics for secure key generation.
• Locally training NNs and transmitting pseudo-keys to enable dynamic key transformation and integrated data transmission.
• Leveraging inconsistencies in channel characteristics between legal parties and eavesdroppers to ensure NN model security.

Main Results:

• Demonstrated feasibility of the proposed system over 100 km with a 100 Gbps bit rate.
• Achieved a key generation rate (KGR) of 50 Gbps for legal parties.
• Realized a key disagreement rate (KDR) of 50% for illegal eavesdroppers, ensuring security.

Conclusions:

• The proposed NN-based system offers a secure and efficient solution for optical transmission, overcoming previous security and performance limitations.
• The system provides robust security against eavesdropping attempts, even when fiber lengths are known, due to unique NN training for legal users.
• Dynamic key transformation and integrated transmission enhance overall system performance and security in optical networks.