The task of seismic phase association is to correlate the onsets of radiated seismic energy with an underlying source. Commonly applied within seismic monitoring networks for event detection, it forms a vital component of many seismic processing pipelines. With the complexity of this task naturally increasing with the number of phases to simultaneously correlate, rapid advancements in the number of sensors per seismic deployment, along with improved picking algorithms have greatly increased the volume of phases now recorded across seismic networks. Although traditional phase association methods work well for historic catalogs, they become unreliable when tasked with associating the frequent smaller events recorded in the latest seismic datasets. Accurately correlating such events is crucial if seismologists are to better understand the underlying physical processes. The phase association problem is, therefore, being revisited with novel techniques now being applied to improve performance. We present a new technique for associating seismic phases, Hyperbolic Event eXtractor (HEX). HEX adapts the logic of Random Sample Consensus, a model estimation approach widely used in the computer vision community and specifically designed to deal with high proportions of noise in the data distribution. We demonstrate the performance of HEX in associating phases over a synthetic dataset for a regional seismic network in northern Chile. Synthetic testing reveals that HEX can correlate seismic phases when events have up to a average spacing.