In VANETs, service requests are often closely related to vehicle locations, and most services require users to share their real-time locations with road side unit. Directly transmitting locations of users to road side unit poses a risk of leaking location privacy. We propose a location privacy protection algorithm based on dynamically adjustable security coefficient for areas with a sparse cluster head and neighbor-tagged vehicles. According to the large-scale data characteristics of VANETs, the algorithm calculates service similarity and the k-value of privacy requirement using vehicle service information from neighbors and the user during the offline phase. In the online phase, the dynamically adjustable security coefficient of tagged vehicles locations is computed using neighbor-tagged vehicle location information to evaluate the security of the locations. Additionally, when neighbor-tagged vehicles fail to meet the user privacy requirement, an elliptic curve division method is adopted to generate fake locations as supplements. Finally, the optimal location is selected as the service request location. Experimental results demonstrate that the proposed algorithm effectively protects user location privacy during service requests, improves service quality, and reduces communication costs.