Nationwide, financial losses due to financial statement frauds (FSF) are mounting. The industry recognizes the problem and is just now starting to act. Although prevention is the best way to reduce frauds, fraudsters are adaptive and will usually find ways to circumvent such measures. Detecting fraud is essential once prevention mechanism has failed. Several data mining algorithms have been developed that allow one to extract relevant knowledge from a large amount of data like fraudulent financial statements to detect FSF. Detecting FSF is a new attempt; thus, several research questions have often being asked: (1) Can FSF be detected? How likely and how to do it? (2) What data features can be used to predict FSF? (3) What kinds of algorithm can be used to detect FSF? (4) How to measure the performance of the detection? And (5) How effective of these algorithms in terms of fraud detection? To help answer these questions, we conduct an extensive review on literatures. We present a generic framework to guide our analysis. Critical issues for FSF detection are identified and discussed. Finally, we share directions for future research.