Research of buffer overflow vulnerabilities detection based on novel K-means clustering

In order to excavate the vulnerabilities of buffer overflow and ensure the safety of the software itself, a kind of buffer overflow vulnerabilities mining method based on novel K-means clustering is proposed. We utilize relevant theories of the buffer to analyze the initial sets of test data collected on the basis of feature similarity and path constraint weights, and then combine with the initial data sequence similarity and path constraint weights to make a higher accuracy and lower non-response rates. This method improves the efficiency of discovering buffer overflow vulnerabilities. The experiment results also show that higher efficiency in the aspect of buffer overflow detection has obtained compared with a series of other experiments. 1553-9105/Copyright © 2015 Binary Information Press