Indoor positioning systems (IPS) based on RSS fingerprints have received significant attention in recent years, but they are unfortunately vulnerable to RSS attacks that cannot be thwarted by conventional cryptographic means. In this paper, we identify two practical RSS attacks on RSS-fingerprint-based IPS (RSS-IPS. In both attacks, the attacker learns the RSS-fingerprint database at the IPS server by acting as a normal user repeatedly issuing location queries and then impersonates selected APs with fake ones under his control. By carefully tuning the locations and transmission power of fake APs, the attacker is able to control the RSS experienced by victim users at target locations, leading to either a large location error or the IPS server misled into returning a fake location of the attacker's choice. We further design a fingerprint-matching mechanism based on a novel truncated distance metric as the countermeasure. Trace-driven simulation studies based on real RSS measurement data demonstrate the severe impact of the proposed attacks and also the effectiveness of our countermeasure.