A modified iterative water-filling algorithm is proposed in which power from only those subcarriers of the near end user are reduced, which has the worst effect on bit rates of far end users. The power back off from these subcarriers is done by reducing power spectral masks at respective subcarriers. The results after simulating the proposed algorithm give significant performance advantages in terms of data rate over the traditional iterative water-filling algorithm. These results approach the performance of highly complex optimal spectrum management algorithms while maintaining the complexity of the traditional iterative water-filling algorithm.