Tor miner and mine pool communication flow detection method based on cell anonymous interaction periodicity characteristics
By extracting the periodic characteristics of anonymous interaction of telecommunications cells in the Tor network and combining with machine learning classifiers, the accuracy and complexity of traffic detection between miners and mining pools in the Tor anonymous network are solved, and efficient traffic identification and monitoring effects are achieved.
Patent Information
- Application Number
- CN202410902529.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-07
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-07-07
AI Technical Summary
The existing Tor miner and mining pool communication traffic detection method has low recognition accuracy when facing the traffic disguise and obfuscation technology of Tor anonymous network, and the traditional method is not applicable to Tor cell slice transmission data, and the high-dimensional traffic feature set leads to data Processing and model training are complex.
Based on the periodic feature extraction method of cell anonymous interaction, the Tor data set is constructed, data preprocessing and timing aggregation are performed, cell response, time and counting features are extracted, feature weights are calculated using LIME, and the machine learning classifier is combined with a machine learning classifier to identify miners and mines. Pool traffic.
It realizes accurate identification of the communication traffic between miners and mining pools in the Tor anonymous network, with a monitoring effect of 99%, simplifying feature extraction and model training, and reducing computing costs.
Smart Images

Figure CN119046747A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of cyberspace security, and in particular relates to a method for detecting communication flow between a Tor miner and a mining pool based on periodic characteristics of anonymous interaction of cells. Background Art
[0002] At present, a large number of research methods have been proposed for detecting the communication behavior between miners and mining pools, which can be mainly divided into endpoint-based and network-based identification methods. The endpoint-based method detects the communication behavior between miners and mining pools by monitoring software operations, website visits or hardware status. It can usually achieve a fairly high accuracy, but it requires the installation of software on each terminal device, which involves user private data and requires regular updates, making it difficult to deploy this method on all devices. The network-based method only needs to be deployed at the network entrance, and the communication behavior can be efficiently detected by analyzing the network traffic generated between miners and mining pools. The communication between miners and mining pools is mainly based on the Stratum protocol. Researchers analyze the communication protocol and traffic characteristics, identify the communication traffic between miners and mining pools through machine learning or rule methods, and further regulate the communication behavior between miners and mining pools.
[0003] Traffic camouflage and obfuscation techniques of anonymous communication networks can hide the communication behavior characteristics between miners and mining pools, further increasing the difficulty of detection. Tor (The Onion Route) is one of the most widely used anonymous communication networks. When users use Tor to access the network on the client side, they need to go through multiple hops to reach the target address, which can effectively hide the information of both parties in communication. In order to further enhance anonymity, Tor is often used in combination with obfuscation techniques. These obfuscation techniques will disguise the traffic between the client and the Tor anonymous network as ordinary traffic or encrypt it, making traffic analysis difficult. Due to the encrypted transmission characteristics of the Tor protocol, traditional analysis methods such as port detection and packet inspection are ineffective for Tor traffic identification.
[0004] At present, the research on Tor miner-pool communication flow detection has the following problems: (1) The traditional miner-pool communication flow detection method is mainly based on TCP transmission extraction features, which is not applicable to the Tor network that uses fixed-size Tor cell slices to transmit data, and Tor uses traffic obfuscation and encryption technology, which also reduces the recognition accuracy of this type of method; (2) The current Tor flow identification method does not specifically conduct in-depth research on the communication behavior between miners and mining pools, does not fully consider the interactive characteristics of communication behavior, uses a high-dimensional flow feature set, and the data processing and model training process are complicated. Therefore, the present invention proposes a Tor miner-pool communication flow detection method based on the periodic characteristics of cell anonymous interaction, combines the Tor anonymous communication mechanism with the miner-pool interaction mechanism, extracts a cell interaction feature set with high discrimination, uses a variety of machine learning models for classification, and finally determines a suitable machine learning classification model based on various evaluation indicators. Summary of the invention
[0005] In order to effectively supervise the communication activities between miners and mining pools in the Tor network, the present invention provides a Tor miner and mining pool communication traffic detection method based on the periodic characteristics of anonymous cell interaction, which can accurately identify the communication traffic between the miner host and the mining pool server in the Tor anonymous network. The method first constructs a Tor data set, including capturing the communication behavior traffic between miners and mining pools encrypted by Tor in the Tor network environment, and collecting public Tor application traffic as background traffic; then performs data preprocessing, including group flow, removing failure cases, and segmenting Tor cells; then divides the traffic into time windows based on time series aggregation, and determines the periodic window according to the time interval and density of the aggregation points; then extracts the communication traffic interaction features with high discrimination in the time window based on the communication interaction mechanism between miners and mining pools in the Tor network, mainly including cell response features, cell time features and cell counting features, and uses the LIME method to calculate the feature weights, further screens the features with high influence on model prediction, and constructs the cell interaction feature set MCI. Finally, the extracted features are input into multiple machine learning classifiers for identification, and the appropriate classification model is determined based on various evaluation indicators.
[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions:
[0007] The present invention provides a method for detecting communication traffic between a Tor miner and a mining pool based on periodic characteristics of anonymous interaction of cells. The system comprises the following steps:
[0008] (1) Construct a Tor traffic data set, including Tor_others and Tor_mining. Tor_mining is the communication traffic between Tor miners and mining pools collected by ourselves, and Tor_others is the background traffic;
[0009] (2) Preprocessing the collected raw traffic data, including grouping flows, removing failure cases, and splitting Tor cells;
[0010] (3) Divide the time window based on time series aggregation, calculate the aggregation point time interval and the number of cells, construct a periodic aggregation point sequence, and thus determine the periodic time window size;
[0011] (4) Based on the communication interaction mechanism between miners and mining pools in the Tor network, we extract highly distinguishable communication traffic interaction features, including response features, time features, and quantity features. Then we use the LIME algorithm to calculate feature weights to further filter out features that have a greater impact on model prediction and form a cell interaction feature set;
[0012] (5) The data set is divided into training set and test set in proportion, the extracted features are input into multiple machine learning classifiers for recognition, and the appropriate classification model is determined based on various evaluation indicators.
[0013] Furthermore, the step (1) specifically includes the following sub-steps:
[0014] Build a universal Tor traffic collection environment and access the Tor network through various methods;
[0015] We selected widely circulated cryptocurrencies in the Tor network environment as research objects, and deployed commonly used cryptocurrency collection software to connect to the mining pool for continuous communication;
[0016] Use traffic collection tools to capture the encrypted communication traffic between Tor miners and mining pools generated during the process;
[0017] Collect public Tor application traffic datasets as background traffic.
[0018] Furthermore, the step (2) specifically includes the following sub-steps:
[0019] (2.1) Data flow division: Tor original traffic is divided into session flows according to five-tuples;
[0020] (2.2) Removing invalid instances of access failure: According to the Tor protocol, when there is a problem with the network status, Tor will update the link and select a new node to establish a virtual circuit. In order to remove the impact of the network status on the data flow, it is necessary to remove invalid instances of access failure according to the Response HTTP status code in the preprocessing stage;
[0021] (2.3) Splitting Tor cells: In Tor, in order to prevent attackers from destroying Tor’s anonymity by extracting features related to data length, data is transmitted in the form of Tor cells of fixed size. In Tor networks with link protocol versions v4 and above, the length of Tor cells is set to 514 bytes. Therefore, in the preprocessing stage, first traverse the data packets of the original session flow, extract the TLS layer data by removing the IP header (bytes) and TCP header (20 bytes), and then split the Tor cells according to the length of 514 bytes to convert the session flow into a Tor unit sequence.
[0022] Furthermore, step (3) specifically includes the following sub-steps:
[0023] (3.1) Cluster the consecutive Tor cells with similar timestamps TCT (Tor Cell Timestamp) and calculate the number of cells in the neighborhood U (U (TCT i ,δ)=(TCT i -δ,TCT i +δ)) timestamps, is defined as the aggregation point AP (Aggregation Point). Where n is the number of Tor cells in the aggregation point, then
[0024]
[0025] (3.2) Calculate the aggregation point density APN (Aggregation Point Number), that is, count the number of Tor cells at the aggregation point.
[0026] (3.3) Assume that periodic communication transmits a similar number of Tor cells at a relatively fixed time interval. Therefore, the aggregation points with similar APNs form a sequence l = {..., AP j-1 ,AP,AP j+1 ,...}
[0027] (3.4) Calculate the time interval between aggregation points in the sequence, defined as API (Aggregation Point Interval), and further calculate the mean of API as the time window T, then T = Avg(API j ),
[0028] Furthermore, step (4) specifically includes the following sub-steps:
[0029] (4.1) Analyze the communication interaction mechanism between miners and mining pools in the Tor network, and observe the interaction behavior between miners and mining pools during the communication process based on the captured Tor miner-miner-mining pool communication traffic. Currently, the most widely used communication protocol between miners and mining pools is the stratum protocol. There is less interaction during the communication process, which mainly consists of four types of messages: computing power submission, task issuance, task submission, and result feedback. Therefore, the data packets of the communication traffic between miners and mining pools are relatively sparse, the time interval between Tor cells is relatively long, and the traffic characteristics show regularity within the time window.
[0030] (4.2) Based on the analysis of (4.1), the interaction pattern characteristics between miners and mining pools within the time window are extracted, including cell response characteristics, cell time characteristics and cell count characteristics, where:
[0031] ① Cell response characteristics: Heartbeat messages are used in the communication between miners or mining pool servers to ensure the stability of the connection. In addition, the data packets of the communication traffic between miners and mining pools are usually small, and a large number of pure response packets (Pure Ack sent) need to be sent during the communication process to confirm, that is, data packets that only contain confirmation information. Therefore, the number of pure response cells and the response characteristics related to the time interval are counted.
[0032] ② Cell time characteristics: The mining pool server hosted by Tor hidden service sets access frequency limit for connection requests from the same IP address, so by observing the collected Tor miner and mining pool communication traffic, it can be found that the data packets in the traffic are relatively sparse, and the time interval between Tor cells is relatively long. Therefore, the time interval and duration related characteristics of the communication traffic between Tor miners and mining pools are distinguishable.
[0033] ③ Cell counting characteristics: Monerod, the Monero blockchain node in the Tor network, reduces the frequency of outbound links to 2 every five minutes, so the miners and the mining pool have less interaction during the communication process, and the frequency of the mining client accessing the mining pool is significantly higher than the response frequency of the mining pool server. Therefore, the characteristics related to the number of uplink and downlink cells in the communication traffic between Tor miners and mining pools are distinguishable, including but not limited to the total number of uplink and downlink cells, the number ratio and other characteristics.
[0034] (4.3) The feature weights are calculated using the LIME (Local Interpretable Model-agnostic Explanations) method. Highly influential features are further selected based on the degree of influence of each feature on model prediction to form the cell interaction feature set MCI.
[0035] The appropriate available features and their corresponding meanings in step (4) are shown in the following table:
[0036]
[0037]
[0038] Furthermore, step (5) specifically includes the following sub-steps:
[0039] (5.1) Divide the dataset into training set and test set according to the ratio of 7:3;
[0040] (5.2) Select the machine learning models commonly used in the research of Tor miners and mining pool communication traffic identification, including classification models such as Native Bayes (NB), K-Nearest Neighbors (KNN), and Support Vector Machine (SVM);
[0041] (5.3) The features extracted based on step (3) are used as model input to accurately identify the communication traffic between Tor miners and mining pools. Precision, Recall, F1-score and Accuracy are used as evaluation indicators to determine the most appropriate classification model.
[0042] Compared with the prior art, the present invention has the following advantages and beneficial effects:
[0043] The present invention can effectively identify the communication traffic between miners and mining pools in the current Tor anonymous communication network, and the monitoring effect accuracy rate reaches 99%. It is convenient for network managers to supervise Tor network traffic and illegal communication behaviors between miners and mining pools.
[0044] The present invention divides the Tor cell as the basic unit of feature statistics, which fully conforms to the characteristic of the Tor protocol that data is transmitted in a fixed size of 514 bytes, and can more accurately analyze the Tor network traffic.
[0045] The present invention determines the periodic time window based on time series aggregation, reduces the redundant information in the data, thereby reducing the dimension of the data, can help identify the periodic patterns and behaviors in the communication traffic between miners and mining pools, and simplifies the subsequent feature extraction steps.
[0046] The present invention combines the Tor anonymous communication mechanism with the miner-to-mining pool interaction mechanism, extracts a highly discriminative cell interaction feature set, has good adaptability for Tor miner-to-mining pool communication traffic identification, and achieves high-precision identification effect with a small number of features, thereby overcoming the shortcomings of the traditional miner-to-mining pool communication traffic detection method based on TCP transmission data that is not applicable to the Tor network and the current Tor traffic identification method that has high feature dimensionality and complex data processing and model training.
[0047] The present invention uses the LIME (Local Interpretable Model-agnostic Explanations) method to calculate feature weights, and further screens high-influence features according to the degree of influence of the features on model prediction, effectively reducing the model training workload and computing costs while ensuring the model recognition accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] Figure 1 Framework diagram of the Tor miner and mining pool communication traffic detection method based on the periodic characteristics of anonymous cell interactions;
[0049] Figure 2 Flowchart of the communication traffic collection between Tor miners and mining pools. DETAILED DESCRIPTION
[0050] The technical solution provided by the present invention will be described in detail below in conjunction with specific embodiments. It should be understood that the following specific implementation methods are only used to illustrate the present invention and are not used to limit the scope of the present invention.
[0051] Embodiment 1: The present invention proposes a method for detecting the communication traffic between Tor miners and mining pools based on the periodic characteristics of anonymous interaction of cells. The classification framework is as follows: Figure 1 As shown in the figure, it is divided into five parts. The first part is to build a Tor data set, which includes building a Tor network environment, capturing the communication behavior traffic between miners and mining pools encrypted by Tor in this environment, and collecting public Tor application traffic as background traffic; the second part is data preprocessing, which includes preprocessing the collected original traffic data, including grouping, removing failure cases, splitting Tor cells and other operations; the third part is to divide the time window based on time series aggregation, which includes calculating the time interval of the aggregation point and the number of Tor cells, building a periodic aggregation point sequence, and determining the periodic time window size. The fourth part is feature extraction, which includes extracting highly discriminative cell interaction features within the time window based on the communication interaction mechanism between miners and mining pools under the Tor network, mainly including cell response features, cell time features and cell count features, and then using the LIME method to calculate feature weights, and further screening features with high influence on model prediction to form a cell interaction feature set; the fifth part is to use a machine learning classification model to identify the communication traffic between Tor miners and mining pools. The specific content includes inputting the extracted features into multiple machine learning classifiers for identification, and determining the appropriate classification model based on various evaluation indicators.
[0052] Specifically, a method for detecting the communication traffic between Tor miners and mining pools based on the periodic characteristics of anonymous interaction of cells includes the following steps:
[0053] (1) Construct a Tor traffic dataset, which includes Tor_others and Tor_mining.
[0054] ①Tor_mining is the self-collected communication traffic between Tor miners and mining pools, which includes the traffic generated by the miner host accessing the ordinary mining pool server (such as pool.suppo rtxmr.com) in the Tor network environment, and the traffic generated by accessing the Tor HS server (such as xmres2r7fiosqduq5oy2avvgtfmr3axsrjvgmfc3vh4yjxdo2pvee6yd.onion);
[0055] ②Tor_others is the background traffic and comes from the ISCXTor2016 public dataset.
[0056] The specific process of this step is as follows:
[0057] (1.1) Build a universal traffic collection environment and build Whonix anonymous operating system through Virtual Box virtualization software, which consists of two virtual machines: Gateway and Workstation. Access the Tor network through three common methods, namely, through SOCKS 5 proxy server and through bridge obfs4 and meek to connect to Tor entry node.
[0058] (1.2) Use the commonly used cryptocurrency collection software XMRig to mine Monero XMR and access the mining pool server through the Tor network. Among them, the XMRig software is installed on the workstation, and Gateway is used as the gateway of the Tor network. Wireshark is deployed here to collect the communication traffic between the miner and the mining pool after the three-layer encryption of the Tor client. The main process is as follows Figure 2 shown.
[0059] (1.3) The mining machine connects to different mining pool servers separately and collects traffic to facilitate data labeling. After each miner's communication with the mining pool, the workstation is reset to the initial state to ensure that the data collected in the previous time will not have residual and combined effects on the data of the next communication behavior.
[0060] (1.4) Collect the original Tor traffic of the ISCXTor2016 public dataset as background traffic, which includes Tor traffic of 8 application types such as browsing, chat, and audio-streaming.
[0061] (2) Perform data preprocessing on the collected raw traffic. The specific process of this step is as follows:
[0062] (2.1) Data flow division: The communication traffic between Tor miners and mining pools is divided into session flows according to five-tuples;
[0063] (2.2) Removing invalid instances of access failure: According to the Tor protocol, when there is a problem with the network status, Tor will update the link and select a new node to establish a virtual circuit. In order to remove the impact of the network status on the data flow, it is necessary to remove invalid instances of access failure according to the Response HTTP status code in the preprocessing stage;
[0064] (2.3) Splitting Tor cells: In Tor, in order to prevent attackers from destroying Tor’s anonymity by extracting features related to data length, data is transmitted in the form of Tor cells of fixed size. In Tor networks with link protocol versions v4 and above, the length of Tor cells is set to 514 bytes. Therefore, in the preprocessing stage, first traverse the data packets of the original session flow, extract the TLS layer data by removing the IP header (bytes) and TCP header (20 bytes), and then split the Tor cells according to the length of 514 bytes, converting the session flow into a Tor cell sequence.
[0065] (3) Dividing the time window based on time series aggregation, specifically including calculating the aggregation point time interval and the number of Tor cells, constructing a periodic aggregation point sequence, and thus determining the periodic time window size.
[0066] The specific process in this step is as follows:
[0067] (3.1) Cluster consecutive Tor cells with similar timestamps (TCT) and calculate the number of consecutive Tor cells in the neighborhood U (TCT i ,δ)=(TCT i -δ,TCT i +δ)) timestamps, defined as the aggregation point AP (Aggregation Point). In this example, δ is 2 seconds. n is the number of Tor cells in the aggregation point, then
[0068]
[0069] (3.2) Calculate the aggregation point density APN (Aggregation Point Number), that is, count the number of Tor cells at the aggregation point.
[0070] (3.3) Assume that periodic communication transmits a similar number of Tor cells at a relatively fixed time interval. Therefore, the aggregation points with similar APNs form a sequence l = {..., AP j-1 ,AP,AP j+1 ,...}
[0071] (3.4) Calculate the time interval between aggregation points in the sequence, defined as API (Aggregation Point Interval), and further calculate the mean of API as the time window T, then T = Avg(API j ), In this example, the time window T is approximately 5 minutes.
[0072] (4) Based on the Tor anonymous communication mechanism and the provisions of the Monero official document on the use of anonymous networks, we further extract the highly distinguishable interaction features between miners and mining pools in the time window. Then, we use the LIME method to calculate the feature weights and further screen the features with high influence on the model prediction to form the cell interaction feature set MCI.
[0073] The specific process in this step is as follows:
[0074] (4.1) Extract the interaction characteristics between Tor miners and mining pools within the time window T, including the number of pure response cells and the cell response characteristics related to the time interval, the cell time characteristics related to the cell time interval and flow duration, and the number of uplink and downlink cells. The specific characteristics are shown in the following table.
[0075]
[0076]
[0077] (4.2) The feature weights are calculated using the LIME method, as shown in the following table. The features with high influence on model prediction are further screened, and the top 6 features with the highest weights are selected to form the cell interaction feature set MCI, including Total Pure Ack, Pure AckInterval Mean, IO Rate, Total Fwd Cells, Total Bwd Cells, and Flow Cells / s.
[0078]
[0079] (5) After screening the features, the Tor traffic dataset is divided into a training set and a test set according to the ratio. A variety of machine learning classification models are used to identify the communication traffic between Tor miners and mining pools, and the most suitable model is selected as the recognition model by comparing the recognition effects.
[0080] The specific process in this step is as follows:
[0081] (5.1) The Tor traffic dataset is divided into training set and test set according to 7:3;
[0082] (5.2) Select a total of 8 machine learning classification models, including Native Bayes, QDA, MLP, Random Forest, IDS, AdaBoost, NearestNeighbors and LSTM, and use the selected 6 features as the input of the model for training and testing;
[0083] (5.3) Using F1-score as the evaluation index, the recognition effects of the eight models are compared, as shown in the following table. Among them, AdaBoost has the highest recognition F1-score, reaching 99.5%, and is the most suitable classification model.
[0084]
[0085]
[0086] The above description is only a partial implementation mode of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and equivalent substitutions can be made without departing from the principles of the present invention. These technical solutions after improvements and equivalent substitutions to the claims of the present invention all fall within the protection scope of the present invention.
Claims
1. A method for detecting the communication traffic between Tor miners and mining pools based on the periodic characteristics of anonymous interaction of cells. It is characterized in that The steps include: (1) Construct a Tor traffic dataset, including Tor_others and Tor_mining. Tor_mining is the communication traffic between Tor miners and mining pools collected by ourselves, and Tor_others is the background traffic. (2) Preprocessing the collected raw traffic data, including grouping flows, removing failure cases, and splitting Tor cells; (3) Divide the time window based on time series aggregation, calculate the aggregation point time interval and the number of Tor cells, construct a periodic aggregation point sequence, and thus determine the periodic time window size; (4) Based on the communication interaction mechanism between miners and mining pools in the Tor network, we extract highly discriminative communication traffic interaction features, including cell response features, cell time features, and cell count features. Then, we use the LIME algorithm to calculate feature weights to further screen features that have a greater impact on model prediction, forming the cell interaction feature set MCI. (5) The data set is divided into training set and test set in proportion, the extracted features are input into multiple machine learning classifiers for recognition, and the appropriate classification model is determined based on various evaluation indicators.
2. According to claim 1, the method for detecting the communication flow between Tor miners and mining pools based on the periodic characteristics of anonymous interaction of cells, It is characterized in that Step (1) specifically includes the following sub-steps: (1.1) Build a universal Tor traffic collection environment and access the Tor network through various methods; (1.2) Select the most widely circulated cryptocurrencies in the Tor network as the research object, and deploy commonly used cryptocurrency collection software to connect to the mining pool for continuous communication; (1.3) Use traffic collection attacks to capture the encrypted communication traffic between Tor miners and mining pools generated during the process; (1.4) Collect public Tor application traffic datasets as background traffic.
3. According to claim 1, the method for detecting the communication flow between Tor miners and mining pools based on the periodic characteristics of anonymous interaction of cells, It is characterized in that Step (2) specifically includes the following sub-steps: (2.1) Data flow division: Tor original traffic is divided into session flows according to five-tuples; (2.2) Removing invalid instances of access failure: According to the Tor protocol, when there is a problem with the network status, Tor will update the link and select a new node to establish a virtual circuit. In order to remove the impact of the network status on the data flow, it is necessary to remove invalid instances of access failure according to the Response HTTP status code in the preprocessing stage; (2.3) Tor cell segmentation: In Tor, in order to prevent attackers from destroying Tor’s anonymity by extracting features related to data length, data is transmitted in the form of Tor cells of a fixed size. In Tor networks with link protocol versions v4 and above, the length of Tor cells is set to 514 bytes. Therefore, in the preprocessing stage, the data packets of the original session stream are first traversed, and the TLS layer data is extracted by removing the IP header (bytes) and TCP header (20 bytes). Then, the Tor cells are segmented according to the length of 514 bytes, and the session stream is converted into a Tor cell sequence.
4. According to claim 1, the method for detecting the communication flow between Tor miners and mining pools based on the periodic characteristics of anonymous interaction of cells, It is characterized in that Step (3) specifically includes the following sub-steps: (3.1) Cluster consecutive Tor cells with similar timestamps (TCT) and calculate the number of consecutive Tor cells in the neighborhood U (TCT i ,δ)=(TCT i -δ,TCT i +δ)) timestamps, defined as the aggregation point AP (Aggregation Point), where n is the number of Tor cells in the aggregation point, then (3.2) Calculate the aggregation point density APN (Aggregation Point Number), that is, count the number of Tor cells at the aggregation point. (3.3) Assuming that periodic communication transmits a similar number of Tor cells at a relatively fixed time interval, the aggregation points with similar APNs form a sequence l = {..., AP j-1 ,AP,AP j+1 ,...} (3.4) Calculate the time interval between aggregation points in the sequence, defined as API (Aggregation Point Interval), and further calculate the mean of API as the time window T, then T = Avg(API j ), 5. According to claim 1, the method for detecting the communication flow between Tor miners and mining pools based on the periodic characteristics of anonymous interaction of cells, It is characterized in that Step (4) specifically includes the following sub-steps: (4.1) Analyze the communication interaction mechanism between miners and mining pools in the Tor network. Based on the captured Tor miner-miner and mining pool communication traffic, observe the interaction between miners and mining pools during the communication process. Currently, the most widely used communication protocol between miners and mining pools is the stratum protocol. There is less interaction during the communication process, which mainly consists of four types of messages: computing power submission, task issuance, task submission, and result feedback. (4.2) Based on the analysis of (4.1), the interaction pattern characteristics between miners and mining pools within the time window are extracted, including cell response characteristics, cell time characteristics and cell count characteristics, where: ① Cell response characteristics: Heartbeat messages are used in the communication between miners or mining pool servers to ensure the stability of the connection. In addition, the data packets of the communication traffic between miners and mining pools are usually small. During the communication process, a large number of pure response packets (Pure Ack sent) need to be sent for confirmation, that is, data packets that only contain confirmation information. Therefore, the number of pure response cells and the response characteristics related to the time interval are counted. ② Cell time characteristics: The mining pool server hosted by Tor hidden service sets access frequency limit for connection requests from the same IP address. Therefore, the collected Tor miner and mining pool communication traffic is found to have sparse data packets and long time intervals between cells. Therefore, the time interval and duration related characteristics of the communication traffic between Tor miners and mining pools are distinguishable. ③ Cell counting feature: The Monero blockchain node monerod in the Tor network reduces the frequency of outbound links to 2 every five minutes, so the miners have less interaction with the mining pool during the communication process, and the frequency of the mining client accessing the mining pool is significantly higher than the response frequency of the mining pool server. Therefore, the characteristics related to the number of uplink and downlink cells in the communication traffic between Tor miners and the mining pool are distinguishable, including the total number of uplink and downlink cells and the number ratio characteristics. (4.3) The feature weights are calculated using the LIME (Local Interpretable Model-agnostic Explanations) method. Highly influential features are further selected based on the degree of influence of each feature on model prediction to form the Mining Cells Interact (MCI) feature set. The suitable available features in step (4) include cell response features, cell time features and cell counting features, wherein the cell response features specifically include the number of pure acknowledgment cells (Total Pure Ack) and the mean time interval of pure acknowledgment cells (Pure Ack Interval Mean); the cell time features specifically include the maximum uplink cell time interval (Fwd IATMax), the mean uplink cell time interval (Fwd IAT Mean), the total uplink cell time interval (Fwd IAT Total), the standard deviation of the uplink cell time interval (Fwd IAT Std), the maximum flow cell time interval (FlowIAT Max), the flow duration (Flow Duration), the mean flow cell time interval (FlowIAT Mean), the minimum flow cell time interval (FlowIAT Min) and the standard deviation of the flow cell time interval (Flow IAT Std); the cell counting features specifically include the total number of uplink cells (Total Fwd Cells), the number of flow cells per second (Flow Cells / s), the total number of downlink cells (TotalBwd Cells) and the ratio of the number of upstream and downstream cells (IO_Rate).
6. According to claim 1, the method for detecting the communication flow between Tor miners and mining pools based on the periodic characteristics of anonymous interaction of cells, It is characterized in that Step (5) specifically includes the following sub-steps: (5.1) Divide the dataset into training set and test set according to the ratio of 7:3; (5.2) Select the machine learning models commonly used in the research of Tor miner and mining pool communication traffic identification, including Native Bayes (NB), K-Nearest Neighbors (KNN), and Support Vector Machine (SVM) classification models; (5.3) The features extracted based on step (4) are used as model input to accurately identify the communication traffic between Tor miners and mining pools. Precision, Recall, F1-score and Accuracy are used as evaluation indicators to determine the most appropriate classification model.
Citation Information
Patent Citations
Data flow windowing and triggering
CN107209673A
Lightweight Tor flow classification method and system based on quaternary feature fusion graph
CN114187485A
Tor site passive fingerprinting system using convolution neural network and method thereof
KR102234698B1
Communications network performance
WO2018167092A1
Cited By
Virtual currency mining machine number identification method and device, electronic equipment and storage medium
CN120934798A
Methods, devices, electronic equipment and storage media for identifying the number of cryptocurrency mining machines
CN120934798B
