In this paper, the noise in anti-eavesdropping system based on acoustic masking is analyzed and the loudness and sharpness which affect the subjective hearing comfort are quantitatively calculated. A new active noise control (ANC) method for anti-eavesdropping system is realized by improving the Filtered-X LMS (FXLMS) algorithm. It can improve the efficiency of active noise cancellation. Experiments show that the proposed method and system can effectively mitigate the redundancy and annoying noise in the speech-protected area. It can significantly reduce the loudness and sharpness of the noise in the protected area and improve the auditory comfort of the talk participant.