In recent years, mobile edge computing and crowdsensing technologies have developed rapidly, with an increasing number of smart devices and sensors being deployed around users to collect and analyze data in real time, thus driving innovations in information processing and application scenarios. However, achieving efficient and real-time risk assessment at the sensor level while ensuring data privacy and security has become a critical research focus. This paper proposes a dynamic risk-adaptive lightweight privacy protection framework for mobile edge crowdsensing scenarios. The framework leverages advanced machine learning algorithms deployed on resource-constrained devices to perform real-time risk assessment on sensor data. Based on the quantified risk results, the system adaptively applies differential privacy mechanisms to protect sensitive data and further encrypts the data prior to transmission to edge nodes. Experimental results demonstrate that the proposed framework not only achieves accurate risk prediction and data analysis on lightweight devices, but also effectively safeguards user data privacy, offering an efficient and feasible solution for the secure management of mobile edge crowdsensing systems.