With the rapid developments of mobile devices and online social networks, users of mobile social networks (MSNs) can easily discover and make new social interactions with others by profiles matching. However, personal profiles usually contain sensitive information of individuals, while the emerging requirement of profile matching in proximity mobile social networks may occasionally leak the sensitive information and hence violate people's privacy. A profile matching protocol in MSNs is proposed, users utilize the confusion matrix transformation algorithm and dot product to achieve secure and efficient matching results; at the same time, users can customize the matching metrics to involve their own matching preference and to make the matching results more precise. In addition, opportunistic computing is adopted to simulate the real friend making senario to guarantee the effectiveness. Security analysis shows that the proposed scheme possesses higher privacy, serviceability, and lower computation and communication cost. Assessed by real social network data, the results demonstrate that the proposed scheme is superior to the existing works.