Conventional generalized sidelobe canceller (GSC) is sensitive to a mismatch between the estimated and actual direction of arrival (DOA) of the desired signal. Such a mismatch induces signal cancellation in the GSC, and it severely degrades the beamforming performance. In this paper, we propose a new decision feedback (DF) technique to increase the robustness against the DOA mismatch. Our new scheme introduces a blind equalizer and a feedback filter in the GSC structure. We first derive Wiener solutions for the DF-GSC with perfectly matched and mismatched DOA and show that the problem of signal cancellation can be avoided. Then, we consider the adaptive GSC implementation in which the least-mean-square (LMS) algorithm is used for weight adaptation. In addition to the improved robustness, the proposed scheme also remedies the slow convergence problem inherent in the conventional adaptive GSC structure. The convergence behavior of the LMS-based DF-GSC is fully analyzed and the analytic signal-to-interference-plus-noise ratio (SINR) is also derived. Finally, simulation results demonstrate that while the proposed structure can considerably enhance the overall performance, it has greatly improved robustness as compared to other existing robust adaptive beamformers.