Maritime piracy has risen in the public awareness as attacks off the coast of Somalia '.have significantly increased since 2007 and, more recently, in West Africa's Gulf of Guinea. Commercially operated ships are at high risk of becoming piracy victims unless robust security measures are taken to prevent such attacks. This thesis attempts to provide a methodology that can be used by marine risk analysts in order to advise a shipowner on the likelihood of a piracy attack against a particular ship at a certain time and under certain environmental conditions. It was also envisaged to make an assessment of the threats that such an attack leads to hijacking of the ship. Due to the highly complex environment a ship is operating in the likelihood of a successful hijacking depends on many factors. These include wind and weather conditions, ship characteristics such as freeboard and speed: the cargo that is being carried, the presence of naval forces in the sea area as well as the security measures taken by the crew. In the current difficult economic climate for ship operations, with low freight rates and tonnage overcapacity for many cargoes, a shipowner is reluctant to invest in unnecessary security measures. Especially the cost of a professional private armed security team is a great expenditure that has to be well thought through and based on a realistic and profound risk analysis in order to balance the risk cost and benefits. This thesis proposes a Bayesian Network model (BN-model) to deal with this multiattribute decision analysis problem under high uncertainty and in such a complex environment. The generic BN-model accommodates all relevant factors that COntribute to a hijacking of a ship. Its appropriateness was tested. Given that some BN-nodes are not based on reliable historical data and, fur t her, that it is difficult to reflect the real-world problem using numerical values, domain experts were requested to provide t heir assessment using linguistic terms. The experts' judgments were processed and aggregated using an enhanced evidential reasoning approach. The weights were determined for the attributes (BN-nodes) by applying the Analytical Hierarchy Process (AHP) and the utility values of the evaluation grades provided for calculating a single crisp number for the likelihood of an attack. The final refinement of the piracy BN-model was achieved by running a MonteCarlo simulation on the root nodes that had been previously improved through the linguistic input from experts. The results of the simulation were then statistically analysed, i.e. statistical tools were applied including a sensitivity analysis. Such an assessment could be used in deciding which measure(s) are most appropriate to mitigate the risk of a hijack for a ship while sailing through the piracy High Risk Area while taking into account the budget constraints mentioned above. This t thesis, comprising of six chapters, follows a step-by-step approach with the aim of developing a commercially viable risk assessment product that, after refinement and conversion into appropriate software, could be used by shipowners, ship operators or charterers to assess the risk for a particular ship when transiting pirate-infested areas using the Bayesian Probabilistic Network approach in conjunct ion with an enhanced Evidential Reasoning methodology, the Analytical Hierarchy Process and Monte-Carlo simulation as underlying methodologies.