Abstract: Most present lossy link identification algorithms do not achieve both high stability and low computation complexity. A novel Reference Path based lossy link Identification algorithm (RPI) is proposed to address this problem. RPI first selects reference paths from the paths passing through a link according to the measured performance differences, and then employs a heuristic method to infer the state of the link based on the correlation of its reference paths. Compared with other algorithms, RPI has high stability because of using the information of measurement data sufficiently. It also achieves low computation complexity due to the heuristic link state inference method. Simulation results show that RPI can identify lossy links more accurately than other algorithms.