Research On Key Node Identification Technology In Tactical Network

As information technology develops,the tactical Internet is gradually formed,network confrontation has become the key to winning modern wars.Generally,it is hoped that the least attack resources will be used to cause significant damage to the enemy's communication system,so how to select the best attack node has become the research focus of network confrontation.The identification of key nodes in tactical networks belongs to a more frontier field,and there have not been many research results in this field.Therefore,from the perspective of third-party reception,this thesis studies the identification of key nodes in the network from two aspects based on the physical layer signals on the channel: one is to get the network topology from the physical layer signals,and then identify the key nodes based on the topology;On the other hand,Use reinforcement learning algorithms to interact with the environment multiple times,and then gradually approach the key moments in the network.this algorithm does not need to obtain the network topology.The main work of this thesis are:(1)An algorithm for identifying key nodes of tactical networks based on topology acquisition is proposed.It firstly identifies the node source of the frame in the physical layer signal based on the radiation source identification algorithm;secondly identifies the MAC protocol of network nodes based on physical layer signals;then designs the topology discovery algorithm based on the node identification and MAC protocol identification;finally identifies key nodes based on Topology in the network.The simulation results show that: if the node identification of the frame in the physical layer signal is completed,the results of the algorithm are consistent with the identification results based on the ideal topology when the signal-to-noise ratio is greater than 5dB.The algorithm provided the new idea for the identification of key nodes.(2)The identification of key nodes based on reinforcement learning was designed,and the effects of ?-greedy,softmax and UCB action selection strategies,packet loss rate,average end-to-end delay,and average routing hop reward criteria on the recognition results were simulated and analyzed.The simulation results show that,one is that the algorithm can identify key nodes of the network without obtaining topology,and the recognition results are consistent with some topology-based recognition algorithms;the second is that the key node recognition results obtained by the algorithm depend on the selection of reward criteria.(3)Channel utilization is used as a reward standard in tactical networks,which solves the problem of difficulty in obtaining rewards in tactical networks.The simulation results show that:if the communication is not frequent in the network,the recognition result of the reinforcement learning algorithm with channel utilization as the reward standard is consistent with the traditional deletion-based shortest path method.(4)Reinforcement learning key node recognition algorithm is improved for identifying key node groups.The improved algorithm expands the action selection space,combines the environmental interaction cost and the environmental feedback as rewards.The set of nodes with the highest degree of joint criticality in the network,called the key node group.The simulation results show that the identification algorithm of the key node group avoids the problem of non-convergence of the single key node identification algorithm in the symmetric network,and the convergence speed is faster under the same network.