Handling Uncertainty in Maritime Incident Assessment: An Approach to Safety-Security Classification

Distinguishing whether a maritime incident stems from safety-related, i.e. accidental, or security-related, , i.e. deliberate, causes is a central challenge, particularly in early phases when available information is incomplete, ambiguous, or contradictory. Although both domains may involve similar observable hazards, they differ in underlying mechanisms, institutional responsibilities, and required response measures. Decision makers must therefore evaluate high-impact situations under considerable epistemic uncertainty. This paper proposes a structured decision-support framework based on Credal Networks, which extend Bayesian Networks by representing probabilities as intervals rather than precise point estimates. The approach is designed to be adaptable to specific maritime contexts. It includes a methodology for identifying and structuring observable indicators using a Human-Technology-Environment categorization, followed by the derivation of causal relationships and the estimation of corresponding credal sets. A representative use case involving a technically impaired vessel near critical infrastructure demonstrates the applicability of the method. The results confirm its feasibility and underline the value of credal networks for uncertainty-aware early-phase maritime incident classification.

Keywords: Maritime Security, Maritime Safety, Credal Networks, Decission Support, Critical Infrastructures.