Disclosed are systems and methods preventing data loss of confidential data in a computer system. The described technique includes searching text data for one or more keywords, and then calculating a density of keywords in the text data based on the one or more keywords that match the text data. The technique classifies the text data as containing confidential data based on whether the density of keywords exceeds a threshold value. If so, the described systems may block the use of the text data containing confidential data in a variety of ways.