Hybrid Trust Model for Identifying Malicious Attacks in Underwater Acoustic Sensor Network

In recent years, underwater acoustic sensor networks (UASNs) have been widely used in ocean information collection, which mainly uses various underwater sensors to collect information. However, UASNs are vulnerable to malicious attacks and network security is vulnerable to threats. The trust model has been proved to be able to effectively detect malicious nodes. However, the existing trust model considers the trust source too single, and it is difficult to comprehensively judge the behavior of nodes, which is prone to misjudgment, and does not identify the types of malicious attacks suffered by nodes after detection. To achieve a more accurate trust evaluation of underwater sensor nodes, we propose a hybrid trust model that can identify malicious attacks on the network. First, trust datasets are established by collecting two types of trust evidence of sensor nodes to ensure the diversity of data sources of the trust model. Then we use random forest (RF) algorithm to build trust model and detect nodes according to trust dataset. According to the detection accuracy, out-of-bag (OOB) error rate, and running time, we choose the most suitable parameters for the trust model. In the final stage, we collect the patterns of malicious attacks on nodes according to the detection results, which is convenient for us to make timely responses and reduce the losses of UASNs due to malicious attacks. The simulation results show that the trust model can effectively detect malicious nodes and attack types in the network and has higher detection accuracy than the existing trust model.