Recently, 3D localization has attracted intense interests with its enabled applications, such as smart healthcare, smart home, and smart robots. Conventional 3D localization technologies require to utilize the special-dedicated infrastructures with a large-scale deployment, which results in high cost. However, in order to accomplish an appealing tradeoff between location accuracy and infrastructure cost, designing and implementing a 3D localization system are still a challenging task to solve. To address this issue, we develop a novel 3D localization system by using commercial off-the-shelf (COTS) WiFi infrastructures at a low cost. First, we design a cost-effective hardware scheme to derive the orientation information for 3D localization. Then, we adopt a fusion method of multiple channel state information measurements to improve multipath resolution and identify the target direct path by the azimuth and elevation angle-of-arrival (AOA) and equivalent time-of-arrival estimations, which can be formulated as subspace-based sparse reconstruction with high accuracy. Finally, according to both azimuth and elevation AOA estimations of the target direct path, the target can be localized in 3D space at decimeter-level. We implement the proposed 3D localization system using COTS WiFi infrastructures and validate its performance with extensive evaluations in two indoor scenarios.