Spread spectrum watermarking (SSW) is one of the most promising watermarking techniques. While SSW is very robust against additive noise, it suffers from an inherent noise. The noise results from host interference on detection of the embedded hidden information. The host interference is one of the main reasons for the relatively low embedding rate in SSW and its reduction is important. In this paper, a new method for use in an SSW detector is proposed to reduce the host interference. The proposed method is based on the law of large numbers and symbol detection of embedded information, instead of bit detection. Since the method is based on the detection of symbols longer than one bit, the detector can employ PN sequences with a larger length, resulting in decrease of the interference. The symbol error rate using this algorithm is calculated employing the maximum a posteriori criterion. Comparison of bit error rates between the modified SSW algorithm and conventional bit detection shows that the modified algorithm is much more robust against the host interference.