Sparse power equalization placement for limiting jamming attack propagation in transparent optical networks

The latest advances in Wavelength Division Multiplexing (WDM) technology are making it possible to build all-optical transparent WDM networks, which are expected to be able to satisfy the rapid growth of today’s capacity demand. However, the transparency of such networks makes them highly vulnerable to deliberate attacks, specifically targeting the physical layer. Physical-layer attacks, such as high-power jamming, can cause severe service disruption or even service denial, enhanced by their capability to propagate through a transparent optical network. Several attack-aware routing and wavelength assignment algorithms have been proposed to reduce the possible disruption caused by high-power jamming attacks. However, even with network planning approaches which take network security, specifically physical-layer attacks, into account, resilience to deliberate attacks in such scenarios remains an issue.In this paper, we propose the use of wavelength-selective attenuators as power equalizers inside network nodes to limit the propagation of high-power jamming attacks. Due to the increased cost of optical switching nodes associated with the addition of power equalizers, we aim at minimizing their number through sparse power equalization placement. We developed a set of greedy algorithms to solve what we call the Power Equalization Placement (PEP) problem with the objective of minimizing the number of power equalizers needed to reduce, to a desired level, the propagation of high-power jamming attacks for a given routing scheme. We further improved upon these results by proposing a GRASP (Greedy Randomized Adaptive Search Procedure) heuristic with a somewhat longer execution time, but with significantly superior results. The performance evaluation results indicate that the proposed GRASP heuristic can achieve the same attack propagation reduction as can be obtained by equipping all nodes with power equalizers by placing them at less than 50% of the nodes on average, potentially yielding significant cost savings.