Recently, great progress has been made in WiFi-based device-free indoor tracking systems. However, due to multipath effect and random noise, the estimated parameter may originate from background clutter rather than the moving target, resulting poor tracking performance. In this paper, we design a WiFi-based device-free tracking system with Probability Data Association (PDA). Firstly, we iteratively estimate multidimensional parameters of all dynamic paths, and only signal power is used to avoid random phase error. Secondly, we treat all subcarriers as multiple snapshots and cluster the estimated parameter to reduce the influence of background clutter. Finally, a PDA method is applied to associate the moving target with ideal parameters and achieve continuous tracking. We extensively validate the proposed system in a typical indoor environment and the experimental result shows that our system achieves a median tracking error of 0.63 m using one single link.