Personal data is sensitive information asset primarily needed to be protected. In order to protect personal data, privacy-protected rules, designs, guidelines, and legal arrangements have been developed so far. Especially, Privacy Impact Assessment methods have been developed with a growing interest in European countries. However, developing information technologies leave these studies insufficient. In this work, a new feature based Privacy Impact Assessment method is proposed for the purpose of protection of personal data. This study focuses on evaluating the privacy impact of data set at attribute level instead of evaluating all of the data which is a general approach to protect data. With the help of calculation at feature level, more sensitive and private personal data parts can be defined and hidden. Data homogeneity method is preferred for privacy impact evaluation calculations. The outcome of this work is data items grouped by privacy impact. According to our proposal, more homogeneous data is more sensitive and its privacy is important. The proposed method is tested on two different data set and the obtained results are analyzed. The most important finding of our work is that attributes that do not appear to be private can be private after combining attributes.