Crosstalk-aware dynamic bandwidth allocation algorithm for confidential lightpath protection in next-generation passive optical networks

Time and wavelength division multiplexed passive optical network (TWDM-PON) is a promising optical access infrastructure for supporting new emerging B5G applications, as it provides broadband capacity and high-rate transmission. With the increasing number of security-sensitive end-users, numerous confidential lightpaths (CLPs) are established over TWDM-PON to carry significant information. However, CLPs are vulnerable to crosstalk attacks due to non-ideal WDM-based devices, typically aimed at eavesdropping on the carried data or even disrupting services by exploiting crosstalk effects in the physical layer. Without optical-electrical-optical conversion of transmitted signals, crosstalk attacks can propagate to the entire network and are difficult to eliminate, severely leading to large-scale information leakage. To mitigate the impact of crosstalk attacks for CLPs, this paper presents a novel crosstalk-aware dynamic bandwidth allocation (CA-DBA) algorithm in the shared uplink of TWDM-PON. We first model the generation and propagation of two types of crosstalk attacks. Based on their characteristics, we propose a crosstalk-aware protection method by introducing the graph neural network (GNN) models to identify the attack probability. Following the protection method, a heuristic algorithm with three differentiated polling schemes is proposed, where the security-priority scheme ensures absolute security regardless of time cost, the time-priority scheme prioritizes meeting latency constraints even experiencing risks, and the time-efficient scheme balances them. Simulation results show that both security-priority and time-efficient schemes could reduce the leakage risk to 0 based on a promising detection performance with up to 100% accuracy using the ChebConv model. In contrast, time-priority would keep at least 6% links at high leakage risk, although it improves the time delay by about 48.04%.