The trustworthiness of nodes in a network is practically a complicated concept and strongly depends on the social relationship among the nodes. For a certain model of trustworthiness, the scheme can be optimized to achieve the highest average secrecy rate based on pre-defined information about possibly malicious nodes. In such a scheme, the secrecy rate is considered as a function of the maximum achievable rate (MAR) at the destination, the MAR at the malicious nodes and the corresponding probabilities. However, in some cases, the trustworthiness model is not available, for example, an engineer does not know if a certain node is maliciously eavesdropping, such a function cannot be derived. Consequently, we need to consider the performance of a scheme in a wide range of trustworthiness levels. To evaluate the performance for such a case, in this paper, we propose a novel concept for comparing these leaking rates which is called leaking rate region. We derive the closed form of leaking rate region for the case of two untrusted relays with multiple antennas at the source and the destination.