Hardware Fingerprint Authentication in Optical Networks Assisted by Anomaly Detection

A hardware fingerprint-based physical layer authentication (PLA) is proposed and demonstrated for optical networks using anomaly detection. To accurately classify the legal devices and effectively detect the rogue devices, the unique fingerprint features are extracted from the received optical spectra of optical transmitters by the linear discriminant analysis (LDA) and then identified by the support vector data description (SVDD) algorithm. The recognition accuracy of 97.8% is experimentally demonstrated based on the optical spectra from 8 transmitters after 20 km standard single-mode fiber transmission, while the average false alarm rate and the miss alarm rate are 2.5% and 0.1%, respectively. The results show the full feasibility and excellent performance of the proposed PLA scheme in security enhancement using hardware fingerprints for optical networks without any threshold selection.

I. Introduction

As data traffic grows dramatically and the accessibility of transparent optical networks increases, optical network security is increasingly important [1]. Many countermeasures against eavesdropping attacks have been proposed using chaotic encryption [2]. Although the data confidentiality is effectively enhanced, they are still susceptible to active attacks. For example, a malicious attacker can access the network and masquerade as a legal user or perform a man-in-the-middle attack [3]. Thus, authentication methods must be employed to ensure the legitimacy of the access devices.