Making new social interactions with other users in vicinity is a crucial service in Proximity-based Mobile Social Networks (PMSNs), where a user can find a best matching friend directly through the Bluetooth/WiFi interfaces built in her mobile device. In existing work for such services, users have to publish their interests to do the matching. However, it conflicts with users' growing privacy concerns about revealing their interests to strangers. To tackle this problem, we propose Weighted Average Similarity (WAS) algorithm, which considers both the number of common interests and the corresponding weights on them, to protect users' privacy without reliance on any Trusted Third Party (TTP). Users set their interests into several priority levels with different weights, then WAS can provide a high level similarity value among these participants without revealing any information about their common interests. The security and computation/communication overhead of our scheme are thoroughly analyzed and evaluated via detailed simulations.