Part of the problem in detecting fraud in accounting and marketing (and because such detection is difficult) is that in many situations scammers are actively trying to hide their behavior. Much of the trust market research or explicitly expects to observe the subjects revealing their behavior; it is likely that fraud cannot spread on its own. Moreover, although certain types of fraud can be detected using parametric statistical models (such as logistic regression) formed on a known sample of fraudulent transactions discovered; there are other types of fraud for which the identification of fraudulent actions is very difficult. An example of fraud that the parametric statistical analysis provided is credit card fraud where the legitimate user can identify fraudulent charges at the end of the billing cycle (or before) or credit card identifies a series of actions using theft of the suspect card.