Social networks play vital roles in our everyday lives. Bypassing traditional media channels, social networks have started to serve as one of the main venues for information dissemination. However, the mechanisms of information propagation, as well as those responsible for message virality in these networks, are still poorly understood. Still, the huge amount of available data or records of user interactions and powerful computing resources enable the use of deep learning in the hunt for better understanding of such networks. In this talk, we will discuss the characteristic of mobile social networks, introduce deep learning techniques, and their application in the investigations of mobile social networks' for efficient and secure operations. As an example, we will detail a selection technique of connection link removal from social networks in order to reduce the spread of rumors. We introduce two unique performance metrics called Average Inverse of Shortest Path Length (AIPL) and Rumor Saturation Rate (RSR) to evaluate how fast messages spread inside social networks. The talk will be concluded with different research directions in applying deep learning in mobile social network studies.