Abstract: A data distributor of an organization has hand over sensitive and confidential data to supposedly trusted third parties (agents) .The data is leaked and is discovered in an unauthorized place. The paper presents a model which includes detecting the guilty agent by the means of data allocation strategy and agent guilt model for guilt probability calculation. . Algorithms proposed for distributing data objects to agents, improves our chances of identifying a leaker. In certain scenario the idea of injecting realistic but fake data in real data is considered for further enhancing guilt probability. The major contribution in this system is to develop a guilt model using fake elimination concept.