Time Dynamic "Allow Listing" For Dropbox

Dropbox is a popular cloud-based file hosting service that is widely recognized for its robust data consolidation and security features, which allow for seamless data access across various locations. Our research is centered on a groundbreaking method to bolster Dropbox’s security measures by minimizing the vulnerability window of network ports. This is achieved through the implementation of a dynamic allow-listing strategy that adapts to user behavior patterns, leveraging the Classification and Regression Tree algorithm to identify peak usage periods.In our investigation, we discovered that implementing a time dynamic allow-listing strategy resulted in adversaries needing four to five times more scan attempts, on average, to uncover an open connection. This posed a significant challenge for attackers, as they were more likely to abandon their attempts if a connection consistently appeared closed. Additionally, our algorithmic findings revealed that our models demonstrated greater precision in predicting patterns for the upcoming week compared to monthly patterns. This disparity was attributed to substantial shifts in user behavior between different weeks of the months, with only minor changes observed within the same month.