چكيده به لاتين
Internet of Things (IoT) is a new technology in which processing devices, intelligent physical objects and humans are integrated and collect, process and exchange data to provide advanced and intelligent services. There is a subset of the IoT called machine-to-machine communication, in which devices are interconnected without human intervention, and can carry out heavy processing with the help of other devices. In these networks, there are many different intelligent objects that automatically and intelligently decide, creating a heterogeneous, distributed, and unknown environment. So, the main challenge is how the devices can trust each other. Here, trust is the belief of a node about the nature of another node in reliable behavior and the provision of quality services. As a result, a trust management system can increase the quality of service. On the other hand, the existence of selfish or malicious nodes disturbs the trust management system and reduces the quality of services and causes the failure of the network.
In some of the previous works, how to rate the quality of the received service is not precisely defined and it is not clear what metrics is used to compute trust. On the other hand, the methods proposed to deal with the on-off attack are only capable of recognizing easy and constant patterns of on-off behavior and are ineffective against an attacker who attacks in a random pattern. The methods for coping with badmouthing attacks also have low efficiency due to not using the recommender's trust for the recommender nodes. These methods have a low performance when both types of attacks exist in the network, because the evaluator node can not recognize that a contradictory recommendation received about the evaluating node is the result of a recommendation from a badmouthing attacker or caused by the evaluating node being an on-off attacker. Hence, in this research, a method for managing trust based on the quality of services is presented which has the ability to mitigate on-off and badmouthing attacks in different scenarios. We also define a separate trust for the recommender nodes, which provides a better and more accurate detection of badmouthing attacks. The results of the simulation show that the proposed method is better than other similar methods in identifying attacks in different scenarios and improving the quality of services.
