Opportunistic Spectrum Access And Routing Security In CR Network

The Cognitive Radio Ad Hoc Networks(CRAHNs) include lots of more challenges than the traditional networks. This paper focuses on the opportunistic spectrum access and the route security.In CRAHNs, a secondary user(SU) searches for spectrum opportunities without a central coordinator, which can be modeled as a Partially Observable Markov Decision Process(POMDP). There is ?Curse of Dimensionality? because the POMDP belongs to the NP Complete. We propose an access scheme that selects channels with Monte Carlo method whose estimation accuracy does not matter with the dimensionality of problem. It is only inversely proportional to the square of problem‘s variance. This paper proposes the simplified-MCTS to search the solution space. This proposed algorithm updates nodes information of the search tree by the additional storage matrix instead of the backpropagation. Meanwhile, the algorithm chooses a shallow rollout tree to complete the part of Monte Carlo simulation.As for routing security, this paper discusses one of the route security problems called the black hole attack. In the network, we can capture route tables to gain a rank sequences by using the FP-Growth, which is a data association rule mining. This simulation shows that the network has a stable rank in a sequence when captured enough route tables. When a malicious node invading the network, a suspicious set consists of nodes which are selected by whether the rank of a node is changed in the sequence. Then, we use the DE-Cusum to distinguish the black hole route and normal one in the suspicious set.The outcomes of opportunistic spectrum access simulation show that the network performances under proposed algorithm are better than ones under the stochastic greedy method, same with the Q-learning and MCTS. However, the time overhead of our algorithm is signicantly less than original MCTS. In the simulation of black hole attacking, we use the NS2 to build a black hole attack scenario with 11 nodes. Simulation results show that the proposed algorithm can reduce much vain detection.