Research On The Performance Analysis Method Of Tor Network Nodes

Tor(The Onion Router)network is an anonymous network which uses multiple proxies,traffic obfuscation and other technologies to hide the relationship and content of the communication between the two parties.Limited by the performance difference of the relay node itself and the defect of the bandwidth measurement method adopted by Tor,the communication performance of the Tor network is not good at present.And users want to ensure the safe,reliable and fast delivery of communication content,so improving the Tor network communication performance has increasingly become the focus of research.For improving Tor network performance,most current researches focus on node performance on bandwidth indicators,and the main focus is on the optimization of routing algorithms.For network services,latency and jitter are also of great significance.In order to optimize the routing method and obtain high-performance anonymous links,this paper proposes a method for calculating node performance and estimating link delay.The details are as follows:First,an optimization-based lightweight node performance measurement method is proposed.This thesis analyzes the factors affecting the performance of Tor anonymous links,constructs a network model based on node delay and jitter performance,and proposes a lightweight node delay and jitter performance estimation method.The factors that affect the performance of anonymous links can be roughly divided into node performance and link performance.We design experiments to analyze the influencing factors,and it is concluded that node performance is the main factor.We measure the delay and jitter of the Tor network,construct the corresponding overdetermined equation system,and then convert it into a least squares model with an L2 regular term,use the optimization method to solve it,and obtain the estimated value of the node performance,and then use the result as According to the path selection,the validity of the solution results is verified.Second,a Tor link delay estimation algorithm based on ensemble learning is proposed.In order to further reduce the link end-to-end delay,through theoretical analysis,the factors affecting the delay of anonymous links are extracted to form the characteristics of anonymous links.Anonymous link features include node features and link features.Node features include node delay features,node geographic location and network location features.In the experiment,the node delay calculated by the optimization method and the delay measured based on the leaky pipe topology are used as the node delay characteristics.The latitude and longitude of the relay node is used as the geographical location feature of the node,and the AS number to which the node belongs is used as the network location feature of the node.Based on the characteristics of relay nodes,the geographical distance and network distance characteristics of links are added to form anonymous link characteristics,which are used as learning characteristics of the Stacking classifier.In order to improve the classification effect,random forest,XGBoost and DNN network are used as the primary classifier,and logistic regression is used as the secondary classifier to train the classifier for link delay estimation.Based on the link delay estimation system for routing,the average delay of anonymous links is significantly improved compared to Tor's default routing algorithm.