We apply deep graph convolutional neural networks for Quality-of-Transmission estimation of unseen network states capturing, apart from other important impairments, the inter-core crosstalk that is prominent in optical networks operating with multicore fibers.

In this work, efficient routing, spectrum, and core allocation (RSCA) techniques are proposed for spatial division multiplexed (SDM) elastic optical networks (EONs) utilizing multicore fibers (MCFs). These techniques provide efficient resource utilization with minimal computational complexity, while also taking into consideration physical layer effects. Initially, a crosstalk-aware approach is presented that accounts for the crosstalk effects in the network that can have detrimental impact on the connections established in the network. This approach is subsequently enhanced with a feedback-based procedure that takes into account all physical layer impairments, aiming to minimize the number of connections that cannot be established in the network due to quality-of-transmission (QoT) considerations.

In this work, routing and spectrum allocation (RSA) algorithms together with network coding (NC) are proposed for elastic optical networks. NC has been used in optical networks for protection against link failures and also in multicasting to improve spectral efficiency. In this work, NC is used to protect confidential connections against eavesdropping attacks. The confidential signals are XOR-ed with other signals at different nodes in their path while transmitted through the network. These signals can be combined either at the source node and/or at intermediate nodes. To implement NC for confidential connections, a set of constraints for the NC problem in addition to the constraints of the RSA problem are incorporated to the algorithms. The combination of signals through network coding significantly increases the security of confidential connections, since an eavesdropper will receive a combination of signals from different connections, making it extremely difficult for the confidential signal to be decrypted. A number of RSA strategies are examined in terms of confidentiality, spectrum utilization, and blocking probability. Performance results demonstrate that network coding provides an additional layer of security for confidential connections with only a small increase in the spectrum usage.

In this work, eavesdropping-aware routing and spectrum allocation (RSA) techniques are proposed for elastic optical networks (EONs) using orthogonal frequency division multiplexing (OFDM). To introduce physical (optical) layer security and protect these networks against eavesdropping attacks, spread spectrum (SS) with signal overlapping techniques are used to encode each requested confidential connection. In order to attain access to the signal and compromise a confidential connection, an eavesdropper will now have to lock on the correct frequency, determine the correct code and symbol sequence amongst copropagated overlapped signals, and decode the signal. Different routing strategies and a novel spectrum allocation technique are proposed in an attempt to add an extra layer of security for confidential connections, while also considering spectrum utilization. Performance results demonstrate that while each confidential connection now requires more spectrum as a result of spreading in the bandwidth domain, the overall network spectrum usage is not increased proportionally to the spreading factor due to the utilization of spectrum overlapping techniques.

In this work, deep graph convolutional neural networks (DGCNN) are applied for estimating the quality-oftransmission (QoT) of unseen network states in elastic optical networks (EONs) in the presence of physical layer impairments (PLIs), including inter-and intra-channel crosstalk (XT). The objective is to find a DGCNN-QoT model that accurately estimates network state feasibility. A network state is considered feasible if the QoT of the in-service lightpaths and of the lightpath under provisioning is sufficient; that is, the DGCNN does not only infer about the feasibility of an unestablished lightpath but also whether the feasibility of the in-service lightpaths will be affected by the establishment of a new lightpath due to XT. As DGCNN model generalization over unseen graphs is known to be negatively affected by the number of possible graphs and their dimensionality, problem uncertainty and complexity is reduced by formulating the QoT estimation problem over subnetwork states, capturing only the spatio-temporal correlations that are relevant to the unestablished lightpath at decision time. DGCNN model accuracy is compared to a state-of-the-art deep neural network (DNN) model trained only over per-lightpath information. It is shown that DGCNN achieves accuracies above 92%, while DNN performs poorly with accuracies as low as 77%, as it fails to infer about the feasibility of in-service connections; an indicator of the importance of explicitly considering during the QoT model training, not only the lightpath patterns, but also the network-state patterns capturing the XT effect. Importantly, it is demonstrated that deep graph learning is a promising approach towards accomplishing this objective.