Machine-learning-based method for fiber-bending eavesdropping detection

In this Letter, we present a scheme for detecting fiber-bending eavesdropping based on feature extraction and machine learning (ML). First, 5-dimensional features from the time-domain signal are extracted from the optical signal, and then a long short-term memory (LSTM) network is applied for eavesdropping and normal event classification. Experimental data are collected from a 60km single-mode fiber transmission link with eavesdropping implemented by a clip-on coupler. Results show that the proposed scheme achieves a 95.83% detection accuracy. Furthermore, since the scheme focuses on the time-domain waveform of the received optical signal, additional devices and a special link design are not required. (c) 2023 Optica Publishing Group

Automated summary

In this letter, a scheme for detecting fiber-bending eavesdropping based on feature extraction and machine learning is proposed. 5-dimensional features are extracted from the optical signal in the time-domain, and a long short-term memory network is used for eavesdropping and normal event classification. Experimental data from a 60km single-mode fiber transmission link with eavesdropping implemented by a clip-on coupler are utilized. The proposed scheme achieves a detection accuracy of 95.83%, and it does not require additional devices and special link design as it focuses on the time-domain waveform of the received optical signal.