Mobile application fingerprinting attacks pose serious privacy threats by identifying the applications used by victims, thereby revealing personal preferences and lifestyle habits. In this paper, we present Cellular-Snooper, the first general and real-time mobile application fingerprinting attack in LTE networks. Cellular-Snooper leverages three novel approaches to improve generality and real-time performance: (1) Using a trace segmentation method to reduce the data collection time. (2) Combining active and passive identity mapping attacks to achieve continuous monitoring. (3) Utilizing a modified GAN to augment the training data. We investigate the feasibility of Cellular-Snooper in a commercial LTE network and achieve an accuracy rate of 90.7% within a 10-s attack. Our work demonstrates the feasibility of implementing real-time privacy attacks without requiring any privileges in LTE networks and provides new insights into the vulnerability of LTE standards to privacy attacks and potential directions for the security of LTE traffic.