A CONTEXT ANALYSIS SCHEME OF DETECTING PERSONAL AND CONFIDENTIAL INFORMATION

The frequent information exchange has been spread widely in the Internet making communication convenience to a huge number of people. However, there are many problems related to stolen personal and confidential (PC) information through the Internet and resources in organizations. In order to support the supervision of PC information, there are many PC texts depending on detection systems. The traditional methods depend on words or a sequence of words without considering the context analysis, and many irrelevant candidates of possible PC information are extracted. This paper presents a new detection scheme for non-PC texts by using context analysis. The heart of the new approach is introducing neglect (NEC) expressions that can cancel the detected PC information. It enables us to reduce extra-detections or human efforts for non-PC texts. Experimental results for context data show that the improvement of the presented method becomes 66.5% compared with the traditional methods. Moreover, the human efforts reduce by about 80% comparing to by using the traditional methods.