Abstract: To cope with the low detection probability and high false alarm probability of Mode S reply signal detection under low SNR conditions, a novel recognition method based on triple correlations, which include detection of the whole signal, matched filtering of single pulse and correlation detection using signal preamble with multiple pulses, is proposed. Firstly, correlation of the whole signal is applied to obtain the signal position and the threshold of preamble, then matched filtering of single pulse is adopted to de-noise for signal and to determine the pulses position. Finally, the pulse amplitude and noise power are estimated and correlation of multiple pulses is applied to recognize the signal preamble. The performance of Mode S reply signal recognition is enhanced via triple correlations in this method. The theoretical derivation of detection probability is given. The simulations are conducted to indicate that detection probability of the proposed method can approach theoretical value under the expected false alarm probability, and to evaluate the performance of the proposed method and conventional methods such as direct matched filtering, pulse front edge detection and correlation detection with normalized baseband, and to illustrate the advantages of the proposed algorithm.