A method for detecting possible fraud in a document includes receiving an electronic image of the document and extracting data from the image that includes a plurality of printed characters. Next, one or more properties of the characters are measured, the properties are compared to at least one corresponding normative property to determine whether anomalies are present, and a determination is made as to whether to provide a fraud alert based on the anomalies, if any. Preferably, the method includes reporting the one or more anomalies, if any, to a classifier engine, wherein the classifier engine determines whether to provide the alert. Also, methods for determining character spacing ands/or acuity are provided.