A Projection Pursuit Based Risk Assessment Method in Mobile Ad Hoc Networks

Establishing high performance cooperation and estimating the nodes´ risk in mobile ad hoc networks (MANETS) is currently fundamental and challenging due to the inherent characteristics of MANETS, such as highly dynamic topology and absence of an effective security mechanism. Trust based assessment methods were recently put forward but presumed restrictions to the data samples or presumed weights for node´s attributes is required. In our paper, PPRA method, short for Projection Pursuit based risk assessment, is proposed to analyze node´s creditability. As projection pursuit turns high-dimensional node properties to low-dimension space, all nodes´ risk levels could be clustered easily and accurately. Projection index, the same to judgment index of clustering consequence, is enacted to reveal different node behaviors. By maximizing projection index through Genetic Algorithm (GA) optimal projection direction is obtained, and then every node´s projection value could be calculated. We prove that relationship between projection value and node risk level is that the larger projection value is, the more credible node is. Finally, the results in one-dimension and two-dimension projection prove our method is more efficient and practical than the traditional methods for risk assessment.