Empirical study included the analysis of the following: 1) 40 sets of fraudulent financial statements and 2) 125 sets of non-fraudulent financial statements (unconditional audit report was issued for the sets of financial statements of these companies). The aim of the research is to distinguish financial ratios, the values of which could indicate the fraud in financial statements. Moreover, the logistic regression model of fraud detection in financial statements has been developed. The research is unique for being the first empirical study of its type in Lithuania.