Smartphone app updates are critical to user security and privacy. New versions may fix important security bugs, which is why users should usually update their apps. However, occasionally apps turn malicious or radically change features in a way users dislike. Users should not necessarily always update in those circumstances, but current update processes are largely automatic. Therefore, it is important to understand user behaviors around updating apps and help them to make security-conscious choices. We conducted two related studies in this area. First, to understand users' current update decisions, we conducted an online survey of user attitudes toward updates. Based on the survey results, we then designed a notification scheme integrating user reviews, which we tested in a field study. Participants installed an Android app that simulated update notifications, enabling us to collect users' update decisions and reactions. We compared the effectiveness of our review-based update notifications with the permission-based notifications. Compared to notifications with permission descriptions only, we found our review-based update notification was more effective at alerting users of invasive or malicious app updates, especially for less trustworthy apps.

No abstract

Abstract. Based on the introduction of the traditional feature weighting algorithm TF-IDF, based on the phenomenon that the eigenvalue extraction is not effective when the text to be classified is not uniform, an improved TF-IDF algorithm is proposed in this paper, which considers the uneven text distribution Inside. The experimental results show that the results obtained by the classification algorithm using the improved algorithm are better than the original algorithm in terms of accuracy and recall and make up for the defects of the original TF-IDF algorithm.

No abstract