Sensors? locations play a critical role in many sensor network applications. A number of techniques have been proposed recently to discover the locations of regular sensors based on a few special nodes called beacon nodes, which are assumed to know their locations (e.g., through GPS receivers or manual configuration). However, none of these techniques can work properly when there are malicious attacks, especially when some of the beacon nodes are compromised. This paper introduces a suite of techniques to detect and remove compromised beacon nodes that supply misleading location information to the regular sensors, aiming at providing secure location discovery services in wireless sensor networks. These techniques start with a simple but effective method to detect malicious beacon signals. To identify malicious beacon nodes and avoid false detection, this paper also presents several techniques to detect replayed beacon signals. This paper then proposes a method to reason about the suspiciousness of each beacon node at the base station based on the detection results collected from beacon nodes, and then revoke malicious beacon nodes accordingly. Finally, this paper provides detailed analysis and simulation to evaluate the proposed techniques. The results show that our techniques are practical and effective in detecting malicious beacon nodes.