Cooperative Jamming Based Physical Layer Security And Encrypted Traffic Classification In Wireless Networks

Embedded,mobile,and wireless cyber-physical systems are becoming ubiquitous and utilized in various applications.The open wireless medium makes the communications among wireless equipment increasingly convenient but also brings more potential security issues.Thus,network security becomes a significant problem in wireless systems.With the development of the computing capacity,the shortcomings of traditional cryptographic security methods appears.The emergence of a large amount of Io T equipment also brings significant impacts to traditional key management schemes.In order to solve the security problems of the wireless network,a careful trade-off between security capability,functionality,and cryptographic primitives must be addressed.For example,resource-constrained nodes should dispose fewer computationally intensive tasks,and the lack of tamper-resistance implies that long-term secrets should not be stored in nodes.Except for the traditional security methods,the inherent randomness in the wireless channels and the positions of the nodes can also provide a strategy for increasing security at the level of the physical layer.This thesis presents a cooperative jamming method used in wireless security.In this method,we consider an environment in which there may exist some eavesdroppers whose positions and quantities are unknown.This means there are unlimited eavesdroppers and they can be extremely close to senders or receivers.The main idea of this strategy is to jam the eavesdroppers by a divide-and-conquer algorithm.In other words,we use senders and receivers to generate the positive jamming signals to solve the security problems caused by the eavesdroppers which are close to the receivers.The method needs to set up some parameters for the network nodes in the cooperative jamming,such as the number of the legitimate nodes which act as jammers,the threshold of both senders and receivers,etc.In practice,a suboptimization approach is employed to get the number of jammers.The basic idea is that we treat the legitimate nodes in the audible region as eavesdroppers.If all the legitimate nodes in audible region are jammed,we regard the communication is secure and stop jamming with a new jammer.Besides,it is difficult to choose proper parameters to realize the highest security rate of confidentiality in different network situations.This thesis proposes a feasible approach which uses a simplified machine learning algorithm to obtain a suboptimal value.To lower the resource consumption of the nodes,we select the Nearest Neighbor Algorithm which has a higher rate of convergence and lower computation complexity.Nevertheless,the Nearest Neighbor Algorithm is likely to provide a locally optimal solution rather than a globally.In order to settle this issue under the premise of decreasing the computation complexity,the thesis adopts a simplified genetic algorithm to obtain the globally optimal solution to get the suboptimal parameter values in the cooperative jamming method.In order to enhance the security of the network communications,we propose a security scheme,traffic classification,at the level of the application layer,which is to classify specific application traffic from mixed network flow.Classifying the encrypted traffic can assist network managers to sort different network applications and protocols so as to realize multiple security purposes like network designing,application optimizing and protocol dispatching among Qo S services.Traffic classification method not only provides better Qo S services but also guarantees the security of the network communications.In the past few years,traffic classification based on machine learning has improved obviously.The machine learning algorithm was proved to be effective and accurate.Even the traffic is encrypted,machine learning also has an outstanding performance in traffic classification.Currently,this classification method could be sorted off supervised machine learning,unsupervised machine learning,and semi-supervised machine learning.This thesis used machine learning to realize the traffic classified scheme: in the process of training classifiers,we build a Markov classifier based on Markov model and a C4.5 classifier to help the Markov classifier;Then in testing processing,we use C4.5 classifier and Markov classifier to identify the traffic and update traffic fingerprints.Evaluations and simulations indicate that cooperative jamming has the capability to enhance the network security which can be employed in the initial connections of the wireless network.It can also improve the wireless network security by employing traffic classification method in application layer.