Trust, the fundamental basis of `cooperation' - one of the most important characteristics for the performance of pervasive ad hoc network - is under serious threat with the emergence of counterfeiting and malicious activity. Several constraints exist in the pervasive computing environment such as a device's memory, battery power and computational capability. This results in a high degree of dependence on other devices within the network to provide services. Along with this is the lack of a secure communication medium which invites both active and passive eavesdroppers. In order to restrict the participation of malicious devices, we propose a trust model which is actually a modified form of the well known LPN (learning parity with noise) based Hopper-Blum (N. Hopper and M. Blum, 2000) protocol. Our new refined model has been named as m-LPN (modified learning parity with noise) which is presented in this paper along with an illustrative example