Internet of things (IoT) is viewed as a set of different application domains sharing a certain number of common functionalities. To be efficient, working on IoT devices should take into consideration their limited capacities in energy, memory and data processing as well as their heterogeneity. Besides the issue of limited node capacities, IoT environment is also confronted with substantial challenges concerning security. Therefore, many researches provide several security mechanisms to reduce attacks on IoT networks. In this paper, we concentrate on excluding malicious nodes from the network based on trust management by assigning a trust level to each node in a cluster-based network and calculating its reputation. We also propose a rehabilitation algorithm to reintegrate excluded nodes and a recommendation algorithm for electing cluster heads (CHs) based on reputation of the other nodes in the network called cluster members (CMs). For better energy consumption and for reducing latency, we based our work on edge computing in order to minimize nodes workload. In fact, data storage and process will be done in bases stations (BSs) instead of remote data centers. Finally, we illustrate these proposed concepts in a case study.