Session Traversal Utilities for Network Address Translator (STUN)-based Traversal Approach Using Port Assignment Prediction Mechanism

This paper presents a STUN-based approach as a way to improve the NAT traversal success rate. As a preliminary step, 50 commercially available NATs are categorized according to the mapping and filtering rules. The port number assignment mechanism for each type of NAT is then deeply probed, and it is found to exert a strong influence on the success rate. On this basis, the combined use of STUN and the probed mechanism is presented as an effective way to improve the success rate. The high performance of this approach is demonstrated by a success rate of 94.36%, compared with 86.6% using STUN and 91.36% using a multi-hole punching counterpart. This approach is expected to be widely applied to peer-to-peer (P2P) communication apps, such as those used in voice and video streaming over IP (V2oIP), Internet of Things (IoT), and many more.