The objectives of this paper are, first to document recent insurance fraud trend and insurance fraud literature, second to introduce the concept and basic rationale of the Link Analysis originated from the Social Network Theory, and finally to evaluate the effectiveness of the analysis method by empirically analyzing real claim data. With strong cracking down efforts by insurers against insurance fraud, insurance consumers complaints appear to increase. In order to control this side effect of the anti-fraud efforts, fraud detection techniques have to be not only effective but also scientifically persuasive. Most of previous literature on insurance fraud detection like Derrig(2002), however, have not explore more scientifically persuasive and more practical 'link analysis' approach but mainly focus on difference kinds of scoring models. Comparing to scoring models using attribute data, the link analysis employing relation data could be more effective and practical in assisting SIU(special investigate unit). Based on the empirical analysis both Emergent Group Algorithm and Step Link Analysis have demonstrated its effectiveness to detecting hard insurance fraud.