A novel neurofuzzy methodology for detecting hazardous material is presented in this paper. This methodology can be used to detect chemical agents at very low level of concentrations. Despite the presence of heavy noise, the presented methodology can enhance the Signal/Noise Ratio (SNR) and increase the confidence level of the material identification decision. The material spectrum is first distinguished from background spectrum by a data validation system; subsequently the spectrum is fitted by a non-linear least square method after appropriate preprocessing and features of peaks are extracted; the features extracted serve as input to the material identification system, and then the decision is made using data fusion techniques. Neural networks and fuzzy logic are used synergistically in this methodology. The test results demonstrated that the current methodology can identify the test material reliably and has a very high detection rate. The false positive rate for this system is less than five percent.