We have observed a large number of 5-15 Hz hydroacoustic arrivals in seafloor pressure data collected in 2009-2010 by the Marine Observations of Anisotropy Near Aoteroa (MOANA) experiment offshore the South Island of New Zealand. Preliminary manual picking revealed that a small number (<10%) of these signals appeared to originate from Antarctica (potentially related to ice disintegration events). Automatic detection of Antarctic signals in MOANA data using frequency-wavenumber (f-k) analysis is precluded in the 5-15 Hz band by spatial aliasing. Instead we have developed a two-stage directional detection scheme based on ambient noise Green’s functions between MOANA stations, which reveal noise directivity in the amplitude asymmetry between forward and reverse lags. For each 20-minute time window of the continuous dataset, we cross-correlate all pressure data, and tabulate windows containing significant energy at the predicted acoustic arrival times between stations and originating from backazimuths consistent with Antarctica. We then beamform each candidate window to estimate signal approach angles. We find that our noise-based approach succeeds in detecting our initial manually-picked events. Back-projecting our detections indicates that many correlate with known iceberg locations from satellite remote sensing studies. We also observe that detection frequency is comparatively lowest in Austral winter, as expected.