A Three Stage Approach to Cybersecurity Management for Logistics

Abstract

The logistics industry is benefiting from the fast-growing cyberspace, but is also increasing its exposure to cyberattacks. Interest in cybersecurity in logistics and supply chain management has grown, but this has not been matched by academic research. This thesis aims to develop a methodology to enhance cybersecurity in an interdependent digital logistics network.

The methodology proposed in this thesis consists of three stages: The precautionary planning stage, the real-time recovery planning stage and the aftermath recovery planning stage. In the precautionary planning stage, this study proposes a novel demon game model against a quantal response (QR) adversary to protect critical assets considering the defending budget and the asset dependency, where the QR adversary can define an attack strategy with biases. Each asset in the solution is represented by its security level indicating its desirability for being protected. Due to the non-convexity of the model, this study proposes a Method of Successive Average heuristic with randomised initial conditions (MSAR) to obtain a promising solution. The efficacy of the proposed heuristic is verified using a hypothetical network after consulting a cybersecurity expert.

Although precautionary strategies are implemented to protect critical assets in a cyber network, a high-level adversary can still penetrate the network and launch attacks inside the organisation. Thus, real-time recovery plays an important role in facing real-time cyberattacks. In the real-time recovery planning stage, this study proposes a novel max-min integer programming model subject to a budget constraint to improve network connectivity of a compromised digital logistics network via maximising algebraic connectivity. Due to the NP-hardness of the model, an optimal solution may not be found in a short time. Thus, several heuristic algorithms, including greedy algorithms, tabu search, and relaxed semidefinite programming (SDP) with rounding, are proposed to find promising solutions. Verification of these heuristic algorithms is achieved by applying them, firstly to a hypothetical network, then to a large scale-free network which mimics a digital logistics network.

When attacks have ceased, recovery measures are initiated to recover the damaged network to its normal state. To speed up the pace of full recovery, resilience plays an important role in recovery. The more resilient the network, the quicker it returns to its normal state after an attack. In the aftermath planning stage, this study proposes a novel max-min mixed integer programming model to improve backbone network resilience by maximising the largest eigenvalue of the associated asymmetric weighted adjacency matrix. Due to the NP-hard nature of the problem, this study proposes an algorithm called LAW (Link Asymmetric Weights) to output a resilient network design. Compared with the enumeration algorithm, the numerical experiments demonstrate the superiority of the proposed algorithm in terms of computation time and solution quality.

In addition, this thesis provides some managerial insights to enhance cybersecurity in logistics management. For example, regular training could improve staff awareness of cyberattacks that could lower the risk of being attacked. Also, the proposed tabu search could help decision makers maintain the compromised network at an acceptable functional state. Last but not least, the proposed LAW algorithm could help a focal organisation identify important new links to improve backbone network resilience when building close relationships with its (potential) third-party service providers.

The proposed methodology is believed to be the first for logistics and supply chains so can potentially serve as a blueprint for other industries, like e-commerce, and governments. It could also be used to mitigate the impacts of other types of risks in logistics systems, smart grids and so on.