With the growing risk of privacy breaches in online social networks, privacy protection has become a key issue. To increase users’ privacy awareness and protect their data, there is a need for a simple and effective method of quantifying privacy risk. A user with a higher privacy risk score is more likely to face a serious privacy breach. In this paper, we propose an effective and reasonable privacy risk scoring method. Our method takes into account the granularity of the shared profile items, combines sensitivity and visibility, and generates a privacy risk score for each user. The calculation of sensitivity and visibility are conducted over a response matrix(R) where each element rij indicates the privacy settings level by user i related to profile item j, and uses improved inverse document frequency (IDF) method to calculate the sensitivity values. Most existing work does not consider profile item granularity. In our study, we define the amount of data shared by users as bytes, classify different granularity levels by one-dimensional clustering, and finally obtain the granularity values using the sigmoid function. With the privacy risk score, users can acquire a more intuitive awareness of their privacy status and then defend it by altering privacy settings or lowering the granularity of shared data. In addition, our experiments analyzing real-world and synthetic datasets demonstrate that our method is capable of effectively assessing user privacy risks in online social networks.