Aiming at the problem that wireless sensor networks（WSNs） nodes are vulnerable to attack but difficult to be evaluated and identified

a detection method based on Markov decision process（MDP） is proposed.The method establishes the MDP model for the network nodes and uses the Option method to layer and simplify.The Q-Learning optimization algorithm is used to iterate the Q values of each layer node until convergence

and finally the optimal detection strategy of the model is solved to determine the nodes of the network.The normal

abnormal state

so as to achieve the purpose of detection.The simulation results show that the simulated annealing technique is introduced into the Q-Learning optimization algorithm to dynamically adjust the greedy strategy and add a layer of learning process.It overcomes the problem of low efficiency and slow convergence when the original algorithm is used to solve the model.The detection method is the same as the same type.Compared with the method

the time-consuming cost of obtaining the optimal strategy is lower

the detection rate is higher

and the false positive rate is lower.