Improved methods to assess an individual's risk of developing cancer, to detect cancers at early stages when they can be treated more effectively, to distinguish between invasive and non-invasive cancers, and to monitor recurrence and response to therapy are required to help doctor treat cancer more effectively. Traditional mammography and DNA Microarrays have been studied for early cancer detection and invasive cancer prediction. However, there is still challenging for detecting early cancer and cancer invasiveness simultaneously. In the paper, we presented a method to discover breast cancer dual-function biomarkers from LC/MS/MS plasma proteome which can discriminate not only cancer from normal breast but also invasive cancer from noninvasive cancer. The training set (Study A) and testing set (Study B) are each from plasma samples of 40 healthy women and 40 women diagnosed with breast cancer. Study A contains 30 invasive cancer samples and 10 non-invasive cancer samples, and Study B contains 23 invasive cancer samples, 8 non-invasive cancer samples, and 9 cancer samples with unknown type. First, we identified from Study A 21 differentially express biomarkers between normal and cancer. Then, we trained a Support Vector Machine with five-fold cross-validation for each combination of 5 out of the 21 biomarkers in the training set. Lastly, we found the optimal combination as best dual-function five biomarker panel. Further pathway analysis showed that the five biomarkers have strong connection with the complement and coagulation cascades pathway. This method can be extended to other cancers for dual-function marker identification. In the future, Multiple Reaction Monitoring (MRM) is planned for validation of these potential dual-function biomarkers.