A multi-path secure transmission method based on link security assessment and fountain code

Abstract

The application discloses a kind of multi-path security transmission methods based on link security evaluation and fountain code.The method comprises: first, establish the malicious traffic detection model RFC based on random forest, network node uses the RFC to detect malicious traffic, and the confidence value of network link is calculated, source node selects path when balancing link confidence value and path length, determines the plurality of safe and reliable paths to destination node using fitness function, source node is fountain encoded to data and uses the plurality of paths transmission encoded package, and destination node can decode to obtain original data when receiving enough number of fountain encoded package.The application combines link security evaluation, fountain encoding and multi-path transmission, improves data delivery rate of data, reduces transmission delay while improving data transmission security, and fully utilizes network resources.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of network communication, in particular to a multi-path secure transmission method based on link security assessment and fountain code. BACKGROUND

[0002] In recent years, data leakage events occur frequently, and most of the data leakage is caused by security problems in network data transmission. The main reasons include: the encryption and decryption algorithm usually requires a large amount of computing resources, and the calculation complexity and node operation cost are high; the rise of intelligent algorithms such as semantic recognition makes it possible to mine and analyze the overall content of the data by only obtaining part of the data; the rapid development of quantum computers and supercomputers makes it possible to crack encrypted data in a short time.

[0003] The avoidance routing protocol prevents data from being eavesdropped or intercepted by malicious nodes by setting a secure transmission path. However, the avoidance routing protocol is highly dependent on the accuracy of malicious node detection, and existing research shows that the best malicious node detection effect can reach 80%, still with 20% uncertainty. Moreover, completely bypassing nodes with security suspicion, the number of hops of data transmission is much larger than the actual optimal routing, resulting in low transmission efficiency and high transmission cost; and the destination node must collect all data packets for transmission to recover the original data, which requires high network reliability. When there are missed malicious nodes, the transmitted data will still be eavesdropped or intercepted by malicious nodes. SUMMARY

[0004] The purpose of the present application is to provide a multi-path secure transmission method based on link security assessment and fountain code, which combines link security assessment, fountain coding and multi-path transmission to improve data transmission security while alleviating problems such as low data delivery rate and large transmission delay.

[0005] In order to achieve the above purpose, the present application provides the following technical scheme:

[0006] A multi-path secure transmission method based on link security assessment and fountain code, the method comprising:

[0007] S1, first establish a malicious traffic detection model RFC based on random forest, the RFC collects network traffic data in real time through a network traffic collection module, and trains and tests the established random forest traffic detection model RFC according to the traffic data.

[0008] S2, the network node detects malicious traffic using the RFC model, and calculates the current confidence D of the link using a sliding window algorithm and a designed decay function according to the proportion of malicious traffic in each historical window. ij (t).

[0009] S3. The source node obtains the path set P of the destination node through routing search. For each path p i it has path confidence D(p i ) and path length L(p i ). Among them, D(p k ) = ∏{D ij}, L(p k ) = ∑e ij , where e ij ∈p k , D ij is the link confidence, and e ij is the link length. When selecting the transmission path, the source node considers the confidence and length of the path in an equalized manner and uses the routing model to select multiple paths as the transmission paths.

[0010] S4. When sending data, the source node performs fountain coding on the data to be sent in advance and transmits the coded packets on the determined multiple paths. The number of coded packets transmitted on each path cannot exceed the decoding threshold λ.

[0011] S5. During the transmission process, partial coded packet loss is allowed. As long as the number of fountain coded packets received by the destination node is greater than the decoding threshold λ, the original data can be decoded.

[0012] Preferably, the steps for establishing the RFC model include: the RFC collects network traffic data in real time through the network traffic collection module and analyzes and processes the characteristics of the collected traffic data, such as the length and quantity of data packets, traffic duration, protocol type, destination IP and port number, etc.; input the traffic sample D = {(x1, y1), (x2, y2), …, (x m , y m )}; determine the number N of decision trees to be constructed, and perform the nth random sampling (n = 1, 2, …, N) from the network traffic data set and the training set. Each time, m samples are collected to obtain a sampled subset Dn containing m samples; a total of N samplings are performed to obtain N sample subsets {D1, D2, …, D N}; use the sampled subset D​​​​​​As preferred, the step of calculating the link confidence degree comprises: detecting the traffic transmitted by each neighbor node, detecting the traffic type using the RFC model, and counting the proportion of malicious traffic in each historical window, and calculating the current confidence degree D of the link using the sliding window algorithm and a designed attenuation function ij (t), wherein

[0014]

[0015] wherein w k is the weight of the kth time window, and is obtained through an attenuation function χ n-k , and GF is the proportion of normal traffic.

[0016] As preferred, the step of routing the model comprises: the routing model designs a fitness function F fit , and calculates the fitness value of each path:

[0017] F fit = Z (α1D(p i )+α2H(p i ))

[0018]

[0019] wherein is a threshold value of the confidence degree, and Z takes a value of 0 when the path confidence degree is lower than the set threshold value. H(p i ) is the reciprocal of the normalized path length L(p i ).

[0020] As preferred, the step of sending the fountain encoding packet comprises: the source node performs fountain encoding on the original data, and dynamically adjusts the number of fountain encoding packets to be sent according to the security level required by the data and the confidence degree of the path, and in the sending process, the fountain encoding packets m1, m2, …, m k are transmitted from the source node to the destination node along multiple selected paths, and the total number of fountain encoding packets to be sent on each path is less than the decoding threshold value λ.

[0021] As preferred, the total number of encoding packets received by the destination node is greater than or equal to the decoding threshold value λ, so that the original data packet can be recovered, and partial encoding packets are allowed to be lost in the transmission process.

[0022] In the above technical solution, the multi-path secure transmission method based on link security evaluation and fountain code provided by the application has the following beneficial effects:

[0023] 1. The traffic detection model RFC based on random forest is established, which can effectively detect the source of malicious traffic, calculate the network link confidence degree, and evaluate the path security.

[0024] 2. When selecting a transmission path, a balance should be struck between path confidence and path length to obtain the actual optimal set of transmission paths, which can improve security and reduce transmission latency.

[0025] 3. Using fountain codes to encode data allows for partial data loss; the destination node can decode the data once it receives a sufficient number of encoded packets. This increases routing opportunities and improves transmission reliability. Attached Figure Description

[0026] Figure 1 A flowchart of RFC model construction in a multipath secure transmission method based on link security assessment and fountain codes provided in an embodiment of the present invention;

[0027] Figure 2 A schematic diagram of a network model is provided in a multipath secure transmission method based on link security assessment and fountain codes according to an embodiment of the present invention.

[0028] Figure 3 A schematic diagram of the transmission process in a multipath secure transmission method based on link security assessment and fountain codes provided in an embodiment of the present invention;

[0029] Figure 4 A transmission diagram is provided in a multipath secure transmission method based on link security assessment and fountain codes, as provided in an embodiment of the present invention. Detailed Implementation

[0030] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0031] Example 1

[0032] like Figure 1 The diagram shows a flowchart of constructing a malicious traffic detection model based on random forest, as described in Example 1 of this invention. The flowchart includes the following steps:

[0033] S1, the RFC collects network traffic data in real time through the network traffic acquisition module, and the traffic data will be processed and saved as a CSV file;

[0034] S2, and analyzes and processes the characteristics of the collected traffic data, such as packet length and quantity, traffic duration, protocol type, destination IP and port number, etc.; features with low discriminative power or excessive feature loss are deleted. Data types are converted, transforming non-numerical feature values ​​in the dataset into numerical data. Furthermore, to balance the influence of each input feature value, the original data is normalized to values ​​in the [0,1] interval, as shown in the following formula:

[0035]

[0036] wherein represents the value obtained by data normalization, x i is the value of the input data, x max is the maximum value in the data variation range, x min is the minimum value in the data variation range.

[0037] In order to improve the construction efficiency of the model and reduce the training time, the correlation values between the features are calculated first, and the highly correlated features are processed. When a pair of feature variables is highly correlated, one of the features can be deleted to reduce the dimension without losing too much information. The features with high importance are selected to construct the model.

[0038] S3, the construction process of the random forest flow classification model RFC is as follows:

[0039] input the flow sample D={(x1,y1),(x2,y2),…,(x m ,y m )}; determine the number of decision trees to be constructed N, and perform the nth random sampling (n=1, 2,…, N) from the training set, collect m times each time, and obtain an individual sample set Dn containing m samples; a total of N samplings are performed to obtain N sample subsets {D1,D2,…,D N}; train the ith decision tree model D i (x) using the sample subset D i , when training the nodes of the decision tree model, a part of the features f (f<F) is randomly selected from all the features F of the sample, and an optimal feature is selected from the selected features for dividing the left and right subtrees of the decision tree. Through recursive operation, the establishment of each decision tree and its nodes is completed, and N decision trees are generated.

[0040] S4, use accuracy, recall rate and F1 score as the evaluation indexes of the model performance. And constantly adjust the parameters of the model, the depth of the decision tree and the number of decision trees, etc.

[0041] As shown in Figure 2 , it is a network model schematic diagram in a multi-path secure transmission method based on link security evaluation and fountain code according to example 1 of the present application. Each node in the network uses the above RFC model to detect the traffic sent by the neighbor, and monitors whether the neighbor node is a malicious node. The proportion of normal traffic in the t time window is calculated by the following formula:

[0042]

[0043] wherein GF(t) is the proportion of normal traffic in the t time window, F all(t) is the total number of traffic through the link in the t time window, F err (t) is the number of malicious traffic through the link in the t time window.

[0044] The attack behavior of malicious nodes can be dynamically changed, and the influence degree of GF of each historical time window on the current link confidence is different, and the influence of GF of the latter time window on the current link confidence should be greater than that of the GF of the previous time window. In order to reasonably evaluate the current security of the link, a time decay function is defined to calculate the weight value of GF of each time window on the current link confidence.

[0045] w k =χ n-k 0<χ<1 1≤k≤n

[0046] Wherein x represents a decay factor, the smaller the value of x, the greater the w k Decay. w k Indicates the weight value of GF of the kth time window, and n indicates the total number of time windows. The confidence of node i to node j is calculated according to the formula: wherein

[0047]

[0048] Wherein w k Is the weight of the kth time window, which is obtained through the decay function, and GF is the proportion of normal traffic.

[0049] As Figure 3 shown, in a multi-path secure transmission method based on link security evaluation and fountain code of example 1 of the application, a transmission flow diagram.

[0050] S1, the source node obtains the path set P of the destination node through searching routing, each path p i Has path confidence D(p i ) and path length L(p i ) identification, wherein D(p k ) = ∏{D ij}, L(p k ) = ∑e ij , e ij ∈p k , D ij Is the link confidence, and e ij Is the link length;

[0051] S2, the source node balances the confidence and path length of the path when selecting the transmission path, and selects the actual optimal path set using the fitness function. The routing model designs a fitness function F fit , to calculate the fitness value of each path, wherein

[0052] F fit = Z (a1D (p i ) + a2H (p i ))

[0053]

[0054] wherein is a threshold value of confidence, when the path confidence is lower than the set threshold value, Z takes the value of 0. H (p i ) is the reciprocal of the normalized path length L (p i );

[0055] S3, the source node encodes the original data into fountain encoding packets m1, m2, …, m k , and dynamically adjusts the number of fountain encoding packets sent according to the security level required by the data and the confidence of the path, stipulating that the total number of fountain encoding packets sent by each path is less than the decoding threshold λ. FC i represents the number of fountain encoding symbols transmitted by path R i , N is the total number of fountain encoding symbols transmitted, and n is the number of paths used for transmission. The allocation of encoding packets is subject to the following conditions:

[0056]

[0057] S4, the fountain encoding packets are transmitted from the source node to the destination node along multiple selected paths, and the destination node can decode the original data when it receives a sufficient number of encoding packets.

[0058] As Figure 4 shown, it is a transmission schematic diagram in a multi-path secure transmission method based on link security assessment and fountain code according to Example 1 of the present application.

[0059] The detection effect of malicious nodes cannot reach 100%, and the missed detection situation may occur. Node a1 in the figure is a malicious node that is not detected, and a1 may intentionally discard part of the data packets to cause a drop attack on the data. The existing method requires that the destination node receives all the data packets sent by the transmission node to be considered as a successful transmission. Once part of the data is intercepted, the data needs to be retransmitted or cannot be transmitted. However, the multi-path secure transmission method based on link security assessment and fountain code proposed in the present application combines link security assessment, fountain encoding and multi-path transmission methods, allows part of the encoding packets to be lost, and the destination node can decode the original data when it receives a sufficient number of encoding packets. The link security assessment can screen reliable paths as transmission paths. Therefore, the transmission method of the present application can increase the data delivery rate and reduce the transmission delay on the premise of improving the data transmission security.

[0060] The above described embodiments are only the description of the preferred ways of the present application, and do not limit the scope of the present application. Without departing from the design spirit of the present application, various modifications and improvements of the technical solutions of the present application made by those skilled in the art shall fall within the protection scope determined by the claims of the present application.

Claims

1. A multipath secure transmission method based on link security assessment and fountain codes, characterized in that, include: S1, nodes in the network use a malicious traffic detection model based on random forest (RFC) to detect malicious traffic, and calculate the link confidence D with neighboring nodes based on the aforementioned RFC and the designed attenuation function. ij ; S2, Before sending data, the source node obtains the path set P to the destination node by searching the route. Each path p i All have path confidence D(p) i ) and path length L(p i The source node considers both the confidence level and path length when selecting a transmission path, and uses a routing model to select multiple paths as transmission paths. S3, when sending data, the source node performs fountain encoding on the data to be sent and transmits the encoded packets on multiple predetermined paths. The number of encoded packets transmitted on each path cannot exceed the decoding threshold λ. S4, during transmission, partial packet loss is allowed. The original data can be decoded as long as the number of fountain-coded packets received by the destination node is greater than the decoding threshold λ. The routing model, when acquiring multiple transmission paths from the source node to the destination node, balances path confidence, path length, and communication quality. The fitness function F is used. fit =Z*{α1D(p i )+α2H(p i )+α3Q(p i )} Calculate the path fitness value, where D(p i H(p) represents the path confidence score. i Q(p) is the normalized value of the inverse of the path length. i The fitness function is used to adjust the weights of path confidence and path length and to calculate the optimal set of transmission paths under the current environment. The path confidence must be greater than a set threshold; otherwise, its fitness value is 0. The path with the higher fitness value is selected as the transmission path. When sending data, the source node performs fountain coding on the data and determines the proportion of fountain-coded packets transmitted on each path based on the above-mentioned path fitness value. The number of coded packets transmitted on each transmission path does not exceed the decoding threshold λ.

2. The multipath secure transmission method based on link security assessment and fountain codes according to claim 1, characterized in that, The RFC model construction steps include: T1. First, establish a malicious traffic detection model RFC based on random forest. The RFC collects network traffic data in real time through a network traffic acquisition module and analyzes and processes the characteristics of the traffic data, such as the length and number of data packets, traffic duration, protocol type, destination IP and port number. T2, the RFC construction process is as follows: Input traffic sample D = {(x1,y1),(x2,y2),…,(x m ,y m Determine the number N of decision trees to be built, and perform the nth random sampling (n = 1, 2, ..., N) from the network traffic dataset and training set, sampling m times each time, to obtain a sampling subset D containing m samples. n A total of N sampling operations are performed to obtain N sample subsets {D1, D2, ..., D...}. N }; using the sampling subset D i Training the i-th decision tree model D i (x), when training the nodes of the decision tree model, a subset of features f are randomly selected from all features F of the sample, and the optimal feature is selected from the selected features to divide the left and right subtrees of the decision tree; through recursive operation, the establishment of each decision tree and its nodes is completed, generating N decision trees; T3. Perform a performance evaluation on the constructed RFC model to ensure that the RFC model can accurately detect malicious traffic.

3. The multipath secure transmission method based on link security assessment and fountain codes according to claim 1, characterized in that, The link confidence calculation steps include: network nodes using the RFC model to detect malicious traffic, counting the number of times malicious traffic occurs within each historical window, and using the sliding window algorithm and a decay function to calculate the current link confidence D. ij (t), where Where w k The weight of the k-th time window is obtained through a decay function, and GF is the proportion of normal traffic.

4. The multipath secure transmission method based on link security assessment and fountain codes according to claim 1, characterized in that, The target node can decode the original data by receiving a number of fountain-encoded packets greater than the decoding threshold λ.

Images