Methods for detecting, associating and locating infrasound events recorded on the global International Monitoring System (IMS) infrasound network are presented. By using likelihood arguments, and reducing the use of empirically determined parameters, our techniques enable us to formally quantify the false alarm rate at both station and network levels, and to calculate confidence areas for event localization. We outline a new association technique that uses graph theory for associating arrivals at multiple spatially separated stations, and perform Monte Carlo simulations to quantify the performance of the scheme under different scenarios. The detection, association and location techniques are applied to 10 large events in the Reviewed Event Bulletin of the Comprehensive Nuclear Test Ban Treaty Organization. Out of 10 events, a total of seven were automatically detected and associated. By analysing the three missed events, we identify improvements that might be made to improve the algorithms.