Sensor localization is one of the big problems when building large scale indoor sensor networks because GPS (Global Positioning System) is unavailable in indoor environments. We are developing ZigLoc, a sensor localization system using WiFi APs (access points) as references, which requires no additional infrastructure [1,2]. In ZigLoc, a sensor node measures RSS (received signal strength) of WiFi AP signals using a ZigBee (IEEE 802.15.4) module. Location of a sensor node is then estimated using fingerprints collected for a WiFi localization system. However, ZigLoc exhibits low accuracy due to the RSS offset derived by ZigBee and WiFi modules. The RSS offset is mainly caused by the channel bandwidth difference. In this paper, we present a differential fingerprinting method to improve localization accuracy. Our key idea is that we focus on RSS difference between WiFi APs. RSS difference between APs should be the same when we measure RSS using either ZigBee or WiFi modules. Differential fingerprinting only relies on RSS difference in fingerprint similarity calculation. We conducted experimental evaluations in a practical environment. The experimental evaluations reveal that ZigLoc accuracy was improved by approximately 26 % using the differential fingerprinting method.