Detection of Guilt Model for Data Leakage

Nowadays the data are distributed for the business purposes through trusted agents. There is a chance for leakage of data and can be identified in unauthorized places. The probability of the leaked information came from one or additional agents should be assessed by the distributor, as hostile having been severally gathered by different means. In this paper we propose information allocation ways (across the agents) that improve the likelihood of distinguishing leakages. These ways do not think about alterations of the free information (e.g., watermarks). In some cases, we will additionally inject “realistic but fake” information records to further improve our possibilities of detection escape and distinguishing the problem.