Buffer overflow vulnerability is a widespread and dangerous security problem. Detecting buffer overflow vulnerability has great research value in information security area. This paper proposes a static comprehensive analytical method for buffer overflow vulnerability detection. Firstly, this method adopts many kinds of static detection tools for detecting the source codes and producing their own detecting reports. Secondly, comprehensive analysis is implemented to evaluate the reliability w eights of detecting tools by training process with detection results, and further optimize the detection results. This training process can improve the efficiency of discovering buffer overflow vulnerabilities with lower rate of omissions and misstatements. The experimental results show that compared with single static detection methods, the rates of both false alert and missed alert decrease significantly.