In recent years, location-based service has been widely used in social networks. However, people's locations or trajectory may be disclosed when they continuously use LBS to retrieve point of interests. The privacy disclosure problem not only restricts the development of LBS, but also reduces the quality of service. Recently, location privacy protection has attracted more and more attention. In this paper, aiming at dealing with the location privacy problem in mobile social network applications, we propose a location privacy protection method for multi-sensitive attributes based on l-diversity privacy protection model, and protect the user's location information in client side and server respectively. On the client side, the decomposition algorithm of minimum distance grouping is used to lighten the location data, which makes the processed data satisfy the l1-diversity principle and upload the data to the server in the form of QIT1 (Quasi-Identifier attribute Table) and ST1 (Sensitive attribute Table) to achieve the initial protection of the user's location data. On the server side, the minimum selection priority strategy is adopted to form the l2-diversity group satisfying the multi-sensitive attributes, and the data is uploaded in the form of QIT2 and ST2 to further protect the user location data (where l1 < l2). The experimental results show that this method not only can effectively protect location privacy data, but also has high data availability.