Network security defense method based on intrusion attack confrontation sample

Abstract

The invention relates to the technical field of network security defense, and discloses a network security defense method based on an intrusion attack confrontation sample, which comprises the following steps: according to fingerprint information of each real node in a target network, newly adding a cheating node with disturbance fingerprint information; obtaining network traffic time sequence data of all nodes in a target network, taking the network traffic time sequence data of real nodes as an original sample, taking the network traffic time sequence data of deception nodes as an adversarial sample, and constructing a target function of an adversarial generative network; dividing deception nodes in a target network into regions according to disturbance types of the deception nodes, randomly adding at least one real node into each region, and selecting any real node from each region as a cluster head node; and the cluster head node selects a defense strategy to resist the attack of an attacker on any node in the region. According to the method, the defense requirements of confrontation samples are met, network node intrusion detection is improved through disturbance design and model parameter optimization, and an effective defense strategy is adopted.

Description

Technical Field

[0001] The present invention relates to the field of network security defense technology, and in particular to a network security defense method based on intrusion attack adversarial samples. Background Art

[0002] In recent years, attackers have used more covert methods to bypass traditional network security defenses, resulting in an increasing number of security incidents. In order to improve the detection accuracy and generalization capabilities of network intrusion detection systems (NIDS), more and more systems have begun to adopt machine learning (ML) and deep learning (DL) technologies. These technologies can learn and model normal network behaviors, thereby effectively identifying abnormal activities and potential threats. With the increasing popularity of deep learning in the field of intrusion detection, it has attracted the attention of attackers, and adversarial attacks against deep learning models have been explored. More and more applications and related platforms have begun to carry deep neural network models. However, the inability of deep neural network models to resist adversarial attacks makes it impossible to widely deploy models in various security fields. Summary of the invention

[0003] The purpose of the present invention is to meet the needs of adversarial sample defense, improve network node intrusion detection through perturbation design and model parameter optimization, and adopt effective defense strategies to provide a network security defense method based on intrusion attack adversarial samples.

[0004] In order to achieve the above-mentioned object of the invention, the embodiment of the present invention provides the following technical solutions:

[0005] The network security defense method based on intrusion attack adversarial samples includes the following steps:

[0006] Step 1: According to the fingerprint information of each real node in the target network, add a deceptive node with disturbed fingerprint information;

[0007] Step 2: Obtain the network traffic time series data of all nodes in the target network, use the network traffic time series data of the real nodes as the original samples, and use the network traffic time series data of the deceptive nodes as the adversarial samples to construct the objective function of the adversarial generation network;

[0008] Step 3: Divide the deceptive nodes in the target network into regions according to their disturbance types, randomly add at least one real node in each region, and select any real node from each region as the cluster head node;

[0009] Step 4: The cluster head node selects a defense strategy to resist the attacker's attack on any node in the area.

[0010] Network security defense based on intrusion attack adversarial samples, including:

[0011] Adversarial sample generation module, used to add deceptive nodes with perturbation fingerprint information and obtain network traffic time series data of real nodes and deceptive nodes;

[0012] Objective function construction module, used to construct the objective function of the GAN model based on network traffic time series data;

[0013] Model parameter optimization module, used to optimize the parameters of the GAN model;

[0014] A region division module is used to divide regions according to the disturbance type of the deceptive node, and randomly add at least one real node in each region, and select any real node from each region as a cluster head node;

[0015] The defense strategy selection module is used to select a defense strategy for the cluster head node to resist the attacker's attack on any node in the area.

[0016] Compared with the prior art, the present invention has the following beneficial effects: the present invention meets the need for adversarial sample defense, improves network node intrusion detection through perturbation design and model parameter optimization, and adopts effective defense strategies. The accuracy of intrusion detection is improved through the profit mechanism, and the effectiveness of defense strategy selection is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 is a flow chart of the method of the present invention;

[0019] Figure 2 is a system block diagram of the present invention;

[0020] Figure 3 Schematic diagram of the target network of the invention. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0022] It should be noted that similar numbers and letters represent similar items in the following figures, so once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance, or implying any such actual relationship or order between these entities or operations. In addition, the terms "connected", "connected", etc. can be directly connected between elements, or indirectly connected via other elements.

[0023] The present invention is achieved through the following technical solutions: Figure 1 As shown, the network security defense method based on the intrusion attack adversarial sample includes the following steps:

[0024] Step 1: According to the fingerprint information of each real node in the target network, a deceptive node with disturbed fingerprint information is added.

[0025] The fingerprint information of the node in the target network is used to represent the identity of a host node. The fingerprint information can include multiple information such as Mac address, open port number, running service, operating system, DNS setting, network protocol, abnormal behavior detection, etc. Assuming that there are n1 real nodes in the original target network, this scheme generates n2 spoofing nodes in the target network. The fingerprint information of the spoofing node also includes Mac address, open port number, running service, operating system, etc., which is the same as that of the real node.

[0026] Unlike the real node, the fingerprint information of the deceptive node is disturbed to form disturbed fingerprint information. As an implementable method, the disturbance added to the deceptive node is a network attack, which may include denial of service attack (DoS / DDoS), backdoor attack (BDAI), vulnerability attack (VAI), phishing (PI), man-in-the-middle attack (MitM), SQL injection attack, cross-site scripting attack (XSS), malware attack, credential theft attack, supply chain attack, etc. For the fingerprint information of a deceptive node, at least one of the above attacks can be added, so that the deceptive node has the ability to confuse the attacker. As another implementable method, the disturbance added to the deceptive node is random noise following the normal distribution. As another implementable method, the disturbance added to the deceptive node is environmental changes and human error operations. For example, environmental changes may include environmental factors such as temperature changes, humidity changes, and electromagnetic interference, which can affect the performance of network nodes. Human error operations may be that the administrator enters incorrect parameters or performs incorrect operations when configuring network nodes. This solution is not limited to the method and type of disturbance added. The ultimate goal is to make the fingerprint information of the deceptive node not the fingerprint information of the real node, and to lure or confuse the attacker.

[0027] A deceptive node is a special host designed by the defender to deceive the attacker. The deceptive node contains the images of all real nodes in the target network and can be disguised as any real node in the target network at any time. The deceptive node can be implemented by a virtualized cloud platform or a high-performance physical machine. Therefore, after designing n2 deceptive nodes, there are a total of n nodes in the target network, n = n1 + n2. Figure 3 As shown, the solid circles in the target network represent real nodes, and the dotted circles represent deceptive nodes.

[0028] Step 2: Obtain the network traffic time series data of all nodes in the target network, use the network traffic time series data of the real nodes as the original samples, and use the network traffic time series data of the deceptive nodes as the adversarial samples to construct the objective function of the adversarial generation network.

[0029] The network traffic time series data of the real node obtained is The network traffic time series data of the deceiving node is: Where τ∈T, T represents the sampling time, m τ Indicates the length of the τth sample data.

[0030] n1 As the original sample, n2 As an adversarial sample, construct the objective function of the adversarial generation network:

[0031]

[0032] The adversarial generative network is based on the GAN model. The generator G tries to minimize the gap between the original sample and the adversarial sample, while the discriminator D greatly improves its ability to judge the original sample and the adversarial sample. )’s true distribution P n1 , and adversarial examples (Right now ) of the adversarial distribution P n2 , the discriminator D judges the input sample From the original sample or the adversarial sample. V(D,G) represents the adversarial function of the generator G and the discriminator D.

[0033] Furthermore, this solution optimizes the parameters of the adversarial generative network so that the GAN model improves the accuracy of the correct result output by the discriminator D. The predicted probability of any node i being disturbed by the GAN model is:

[0034] q i (t) = (1-β t )·q i (t-1)+β t ·e i (t-1);

[0035] Among them, q i (t) represents the predicted probability of node i being disturbed in the tth iteration output of the GAN model; q i (t-1) represents the predicted probability of node i being disturbed in the t-1th iteration output of the GAN model; e i (t-1) represents the true value of the disturbance of node i in the t-1th iteration. Since node i is either disturbed or not disturbed, the true value e i (t-1) is a binary classification problem. If node i is not disturbed, then e i (t-1) = 1, if node i is disturbed, then e i (t-1) = 0; β t represents the perturbation factor of the tth iteration, when e i When (t-1) = 0, β t ·e i (t-1) = 0, at this time only relying on (1-β t )·q i (t-1) to improve q i (t) accuracy, so β t Designed for:

[0036]

[0037] Among them, β t-1represents the perturbation factor of the t-1th iteration; ε represents the learning rate; ρ is a constant to avoid the denominator being 0, and ρ=10 can be taken -8 ;μ t is the gradient function of the tth iteration, This formula represents the parameter θ of the adversarial function V(D,G) on the discriminator D at the tth iteration D The purpose is to make the adversarial function V(D,G) converge faster and more accurately.

[0038] If the predicted probability q i The closer (t) is to 1, the higher the possibility that node i is disturbed. i The closer (t) is to 0, the higher the possibility that node i is not disturbed. A threshold can be set according to the actual situation to determine whether node i is disturbed. This solution does not limit how to set the threshold.

[0039] If the output of the GAN model is wrong, the reward given is 0. If the output of the GAN model is correct, the reward given is:

[0040]

[0041] Among them, V i represents the benefit obtained by the GAN model for the prediction result of node i output; γ i represents the profit offset for node i; t represents the current iteration number; t max Indicates the maximum number of iterations of the network traffic time series data of node i; represents the revenue weight; δ represents the learning rate.

[0042] Step 3: Divide the deceptive nodes in the target network into regions according to their disturbance types, randomly add at least one real node in each region, and select any real node from each region as the cluster head node.

[0043] For example, when the disturbance added to the deceptive node is a network attack, the n2 deceptive nodes can be divided into M areas according to the type of network attack, and then the n1 real nodes are randomly assigned to the M areas, and there is at least 1 real node in each area. Generally speaking, the number of deceptive nodes n2 will be less than the number of real nodes n1, so the number of areas M is also less than the number of real nodes n1, which can ensure that each area can be assigned to at least 1 real node. As a preferred solution, the number of real nodes in each area is as close as possible, but not necessarily exactly the same.

[0044] Then, a real node is selected from each area as the cluster head node, which is used to implement attack defense on the nodes in the entire area. When selecting the cluster head node, the network traffic time series data of the real node is first obtained, and the network traffic time series data is converted into traffic features to construct the traffic adjacency matrix G:

[0045]

[0046] Among them, g ij represents the traffic consumed by the i-th node to transmit data to the j-th node in the area; m represents the number of real nodes in the area, i = 1, 2, ... m, j = 1, 2, ..., m; g ii Indicates the traffic consumed by the i-th node to process data.

[0047] Calculate the fitness value of node i:

[0048]

[0049] Among them, E i Represents the fitness value of node i; α 1 , α 2 , α 3 All are weights; g i represents the current traffic of node i, g i0 represents the initial flow of node i; D i_base represents the distance from node i to the base station; D max_base Indicates the maximum distance from the node in the area to the base station; g max Represents the maximum element in the traffic adjacency matrix G.

[0050] The node with the largest fitness value in the area is selected as the cluster head node.

[0051] Step 4: The cluster head node selects a defense strategy to resist the attacker's attack on any node in the area.

[0052] At time t`, the cluster head node selects the defense strategy s k The probability is:

[0053]

[0054] in, Indicates that the cluster head node selects defense strategy s at time t` k The probability of Indicates that the cluster head node selects defense strategy s at time t` k The weight of Indicates that the cluster head node selects the defense strategy s at time t`-1 k The weight of; σ represents the full weight impact factor; Indicates the cluster head node selects the defense strategy s k The expected probability of k represents the kth defense strategy, k = 1, 2, ..., K, and K represents the total number of defense strategies.

[0055] The defensive benefits are:

[0056]

[0057] Among them, R A represents defensive gains; Indicates the cluster head node selects the defense strategy s k The probability of l represents the lth attack method, l=1,2,...,L, L represents the total number of attack methods; r(f l ,s k ) indicates that the cluster head node selects the defense strategy s k Means of resistance to attack l The income obtained when Z defense (s k ) indicates the selection of defense strategy s k In order to reduce the cost resources available for defense measures, the following constraints are imposed on the defender:

[0058]

[0059] Among them, ω d represents the minimum cost utilization, ω d <1; represents the defense strategy vector; y i represents the degree of cluster head node i; γ d represents the node degree coefficient; Z all (s k ) indicates that all real nodes in the area select defense strategy s when acting as cluster head nodes k The total cost of d represents the cost constraint coefficient; i represents the i-th node in the area as the cluster head node, i = 1, 2, ..., m, and m is the number of real nodes in the area.

[0060] like Figure 2 As shown in the figure, this scheme also proposes network security defense based on intrusion attack adversarial samples, including:

[0061] Adversarial sample generation module, used to add deceptive nodes with perturbation fingerprint information and obtain network traffic time series data of real nodes and deceptive nodes;

[0062] Objective function construction module, used to construct the objective function of the GAN model based on network traffic time series data;

[0063] Model parameter optimization module, used to optimize the parameters of the GAN model;

[0064] A region division module is used to divide regions according to the disturbance type of the deceptive node, and randomly add at least one real node in each region, and select any real node from each region as a cluster head node;

[0065] The defense strategy selection module is used to select a defense strategy for the cluster head node to resist the attacker's attack on any node in the area.

[0066] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A network security defense method based on intrusion attack adversarial samples, characterized by: The following steps are involved: Step 1: According to the fingerprint information of each real node in the target network, add a deceptive node with disturbed fingerprint information; Step 2: Obtain the network traffic time series data of all nodes in the target network, use the network traffic time series data of the real nodes as the original samples, and use the network traffic time series data of the deceptive nodes as the adversarial samples to construct the objective function of the adversarial generation network; Step 3: Divide the deceptive nodes in the target network into regions according to their disturbance types, randomly add at least one real node in each region, and select any real node from each region as the cluster head node; Step 4: The cluster head node selects a defense strategy to resist the attacker's attack on any node in the area.

2. The network security defense method based on intrusion attack adversarial samples according to claim 1 is characterized by: In step 1, the fingerprint information includes Mac address, open port number, running services, operating system, DNS settings, network protocol, and abnormal behavior detection; Perturbing fingerprint information means adding disturbances to the fingerprint information of the deceptive node. The disturbances include network attacks, random noise, environmental changes, and human error operations.

3. The network security defense method based on intrusion attack adversarial samples according to claim 1 is characterized in that: The step 2 specifically includes the following steps: The network traffic time series data of the real node obtained is The network traffic time series data of the deceiving node is: Where τ∈T, T represents the sampling time, m τ Indicates the length of the τth sampling data; n1 As the original sample, n2 As adversarial samples, n1 is the number of real nodes, n2 is the number of deceptive nodes, and the objective function of the adversarial generation network is constructed: The adversarial generative network is based on the GAN model. The generator G tries to minimize the gap between the original sample and the adversarial sample, while the discriminator D greatly improves its ability to judge the original sample and the adversarial sample. )’s true distribution P n1 , and adversarial examples (Right now ) of the adversarial distribution P n2 , the discriminator D judges the input sample From the original sample or the adversarial sample. V(D,G) represents the adversarial function of the generator G and the discriminator D.

4. The network security defense method 3 based on intrusion attack countermeasure samples according to claim 1 is characterized in that: The step 2 further comprises the following steps: The predicted probability of any node i being disturbed by the GAN model output is: q i (t)=(1-β t )·q i (t-1)+β t ·e i (t-1); Among them, q i (t) represents the predicted probability of node i being disturbed in the tth iteration output of the GAN model; q i (t-1) represents the predicted probability of node i being disturbed in the t-1th iteration output of the GAN model; e i (t-1) represents the true value of the disturbance of node i in the t-1th iteration. If node i is not disturbed, then e i (t-1) = 1, if node i is disturbed, then e i (t-1) = 0; β t represents the perturbation factor of the t-th iteration, which is: Among them, β t-1 represents the perturbation factor of the t-1th iteration; ε represents the learning rate; ρ is a constant to avoid the denominator being 0, and ρ=10 can be taken -8 ;μ t is the gradient function of the tth iteration, This formula represents the parameter θ of the adversarial function V(D,G) on the discriminator D at the tth iteration D gradient.

5. The network security defense method based on intrusion attack adversarial samples according to claim 1 is characterized in that: In step 3, the step of selecting any real node from each area as a cluster head node includes: Obtain the network traffic time series data of real nodes, convert the network traffic time series data into traffic features, and construct the traffic adjacency matrix G: Among them, g ij represents the traffic consumed by the i-th node to transmit data to the j-th node in the area; m represents the number of real nodes in the area, i = 1, 2, ... m, j = 1, 2, ..., m; g ii It represents the traffic consumed by the i-th node to process data; Calculate the fitness value of node i: Among them, E i represents the fitness value of node i; α1, α2, α3 are all weights; g i represents the current traffic of node i, g i0 represents the initial flow of node i; D i_base represents the distance from node i to the base station; D max_base Indicates the maximum distance from the node in the area to the base station; g max Represents the maximum element in the traffic adjacency matrix G; The node with the largest fitness value in the area is selected as the cluster head node.

6. The network security defense method based on intrusion attack adversarial samples according to claim 5 is characterized by: The step 4 specifically comprises the following steps: At time t`, the cluster head node selects the defense strategy s k The probability is: in, Indicates that the cluster head node selects defense strategy s at time t` k The probability of Indicates that the cluster head node selects defense strategy s at time t` k The weight of Indicates that the cluster head node selects defense strategy s at time t`-1 k The weight of; σ represents the full weight impact factor; Indicates the cluster head node selects the defense strategy s k The expected probability of k represents the kth defense strategy, k = 1, 2, ..., K, and K represents the total number of defense strategies.

7. The network security defense method based on intrusion attack adversarial samples according to claim 6 is characterized by: The step 4 also includes the following steps: The defensive benefits are: Among them, R A represents defensive gains; Indicates the cluster head node selects the defense strategy s k The probability of l represents the lth attack method, l=1,2,...,L, L represents the total number of attack methods; r(f l ,s k ) indicates that the cluster head node selects the defense strategy s k Means of resistance to attack l The income obtained when Z defense (s k ) indicates the selection of defense strategy s k Costs; Select the defense strategy with the highest defense benefit value to resist the attacker's attack on any node in this area.

Images