Many regions around the world are prone to tsunami risk, and their populations are expected to increase. Moreover, past events like the 2011 Great Eastern Japan Earthquake and Tsunami resulted in numerous fatalities and the failure of many coastal protection structures. These events underscore the urgent need for further research in tsunami engineering and the mitigation of tsunami risk. In fact, the capability of coastal protection structures, such as breakwaters, to withstand tsunami loads is not yet fully understood. Additionally, these structures are typically designed to resist wind waves loads rather than the longer waves produced by tsunamis. This research provides further insight into the interaction between tsunamis and composite breakwaters by analyzing experimental results where a unique technique capable of accurately reproducing tsunamis as scaled N and E-waves was used. The experiments employed a 2-dimensional flume where waves were generated with a tsunami simulator and propagated until they impacted a composite breakwater model, inspired by the world-record breakwater in the Kamaishi bay in Japan. This research thoroughly analyzed the results of one of the experiments conducted at the HR Wallingford research center in the UK, with the objective of understanding the response of a composite breakwater when impacted by a tsunami, focusing particularly on the caisson on top of the structure and its stability. Another objective was the development of a numerical model based on the coupling of the two software OceanWave3D and OpenFOAM, aiming to reproduce physical experiments on tsunami-structure interaction and provide further insights beyond the capabilities of physical tests. The results of this research indicate that the pressures and forces induced by a tsunami on the caisson of a composite breakwater have a dominant hydrostatic contribution. The absence of wave breaking as the tsunami approaches the breakwater and shoals on its rubble mound prevents the generation of impulsive forces on the structure. The analysis also shows that, for the considered tsunami at prototype scale, the caisson would be unstable and fail due to sliding, primarily because of the water level and pressure differences on the two sides of the structure. The model developed in this research demonstrated a good accuracy in representing the physical experiment, as evidenced by elevation and pressure time series, with minor limitations on the lee side of the structure and on its rubble mound. With further validation using additional experimental results, this model can serve as a starting point for future studies on tsunami-structure interaction. It overcomes some limitations of physical testing, potentially provides more accurate results for caisson stability analysis, and offers a cost-effective alternative to physical experiments.