Abstract:

The existing trust models for wireless sensor networks seems imprecise when the node credibility presents periodic changes which are affected by nodes’ performance and environment. For that, a subjective trust model based on mixed Gaussian distribution (MGSRM) is proposed. This model builds multiple Gaussian distributions corresponding to different “state”, improves dynamic adaptability of trust values. Current trust value and integrated trust value are separately used to reflect the node credibility under long-term and short-term behavior. A better focus and practicability is obtained. Simulation shows that, compared to existing trust models, MGSRM model has obvious advantages in accuracy, dynamic adaptability and learning ability.

Key words: sensor networks, subjective trust model, mixed Gaussian distribution