Capacity depletion type DDoS intrusion detection and mitigation method based on online time sequence analysis

By deploying detection and mitigation methods based on online timing analysis on programmable switches at the network edge, using the data plane and the control plane interaction, the problems of poor detection and low resource utilization in the prior art are solved, and capacity exhaustion is achieved Efficient detection and mitigation of type DDoS intrusion.

CN119094223APending Publication Date: 2024-12-06HUNAN UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411292728.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-14
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

Existing capacity-depleted DDoS intrusion detection and mitigation methods are not robust when deploying data planes, complex processing logic leads to reduced throughput, and the control plane does not rationally utilize data plane resources, resulting in excessive communication overhead and difficult to achieve high resource utilization. rate and high performance detection and mitigation.

Method used

Deploy detection and mitigation methods based on online timing analysis on the programmable switch at the network edge. Using data plane computing capabilities and comprehensive control of the control plane, collect network status information through dynamic Sketch module, combine information entropy calculation and chaos theory analysis to achieve two stages Detect and mitigate through graph modeling and flow table rule insertion.

Benefits of technology

Real-time detection and mitigation of capacity-depleted DDoS intrusion is realized, system overhead and switch resource consumption are reduced, detection effect and resource utilization are improved, and low time delay and timely intrusion detection and mitigation are achieved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119094223A_ABST
    Figure CN119094223A_ABST
Patent Text Reader

Abstract

The invention discloses a capacity depletion type DDoS intrusion detection and mitigation method based on online time sequence analysis, and belongs to the field of computer network security. The method comprises the following steps: designing a dynamic Sketch activation module, collecting network state information features by using Sketch, and calculating information entropy as a reflection of a network state; preliminary judgment is carried out in a data plane according to the information entropy obtained through calculation, if a suspicious flow is judged, the suspicious flow is reported to a controller through digest, online time sequence analysis is carried out, the state of the suspicious flow is further deduced through phase-space reconstruction and detrending fluctuation analysis, and therefore the final judgment of the network state is obtained; after intrusion judgment, reading network topology and performing graph modeling in the controller, performing IP traceability by using reverse depth-first search, and then inserting a filterable flow table rule to relieve; according to the method provided by the invention, the capacity depletion type DDoS intrusion can be detected and relieved in time, and the method is a feasible and effective method for detecting and relieving the capacity depletion type DDoS intrusion.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of computer network security, and in particular relates to a capacity exhaustion type DDoS intrusion detection and mitigation method based on online timing analysis. Background Art

[0002] Denial of Service (DoS) attack is an attack method that exhausts the target system's resources (such as bandwidth capacity, computing power or memory) by sending a large amount of useless data or requests to it, so that it cannot handle normal traffic or requests through capacity exhaustion, protocol abuse, application layer attacks and logical attacks. Its main purpose is to prevent legitimate users from accessing services. DoS attacks are usually launched from a single source (a device or an Internet connection). Distributed Denial of Service (DDoS) attack is an extended form of Denial of Service attack, which uses multiple infected devices to jointly launch the attack. These devices may be computers, servers, IoT devices, etc. around the world. This type of attack is also one of the main attacks facing cyberspace security in today's era.

[0003] Volumetric DDoS attacks, also known as volumetric attacks, are a type of attack that aims to consume the target network bandwidth capacity or server resource capacity, causing normal traffic to be unable to reach the target. This type of attack floods the target with a large number of data packets or requests, exceeding its processing capacity, causing network congestion or system resource exhaustion, making it impossible for legitimate users to access services.

[0004] Among the detection and mitigation methods for capacity-exhaustion DDoS intrusions that have been proposed, the methods that are mainly deployed on the data plane cannot achieve robust detection, and the overly complex processing logic will reduce the throughput of programmable switches to an unacceptable level; the methods that are mainly deployed on the control plane do not reasonably utilize the limited but effective computing resources in the data plane, and often cause excessive communication overhead between the control plane and the data plane. Therefore, how to make full use of the interaction between the two to achieve a high resource utilization and high-performance detection and mitigation system is still a major challenge at present.

[0005] Under the actual needs, the programmability of the network has been gradually discovered and developed. At the same time, the concept of programmable data plane has been proposed and experimental prototypes have been given for actual needs. Among them, the most widely used tool for realizing network programmability is the Programming Protocol-Independent Packet Processors (P4) language, which defines the processing method of data packets, thereby providing a flexible and protocol-independent way to control and define network devices to achieve more efficient network-related functions. At the same time, P4 uses the P4Runtime API to realize the communication between the control plane and the data plane, and controls it by inserting entries in the match-action table, so that the controller can use other programming languages ​​to cooperate with the programmable data plane for further functional implementation.

[0006] The present invention proposes a capacity exhaustion type DDoS intrusion detection and mitigation method based on online time series analysis for capacity exhaustion type DDoS intrusion. The method deploys the detection and mitigation system on the network edge programmable switch, fully utilizing the limited but effective computing power provided by the data plane and the comprehensive regulation capability of the control plane, and performs comprehensive detection and traceability mitigation based on time series analysis for each target IP address at the edge of the network. For each target IP address at the edge of the network, within the data plane, based on the last time window information, a module called by a dynamic Sketch is used to collect network status information, wherein the module uses a Count-Min Sketch to collect the number of data packets for the target IP address, and uses HyperLogLog Sketch is used to collect the source IP cardinality corresponding to the data packets for the target IP address, so as to measure and estimate the network status. At the same time, using these two network status features, the information entropy calculation form is adopted in the data plane, and the information entropy is used as a comprehensive reflection of the network status. The information entropy calculation result is compared with the easy-to-distinguish attack judgment threshold θ1 and the complex judgment threshold θ2 of the suspicious state in the data plane for preliminary judgment. If it exceeds θ1, the mitigation mechanism is directly triggered through digest. If it does not exceed θ1 but exceeds θ2, the control plane is triggered to further analyze and judge the suspicious state. In the control plane, for each target IP address, a time window is used as a division, and a time series with a fixed length and a value of the information entropy in the time window is maintained. The final judgment of the suspicious state is made through the phase space reconstruction and detrending fluctuation analysis method based on chaos theory. If it is judged as an intrusion, the mitigation mechanism is directly triggered in the control plane. In terms of mitigation, the control plane reads the network topology, models the graph, and uses the flow table rules as the edges of the graph to trace back to the source, finds the source IP address of the intrusion, and inserts the flow table rules for filtering the source IP address to achieve intrusion mitigation. The method proposed in the present invention can realize real-time detection and mitigation of capacity exhaustion type DDoS intrusion. The method makes full use of the interaction between the data plane and the control plane, balances resource consumption and detection effect, and realizes low time delay, effective and timely detection and mitigation of capacity exhaustion type DDoS intrusion. Summary of the invention

[0007] The present invention proposes a capacity exhaustion DDoS intrusion detection and mitigation method based on online timing analysis for capacity exhaustion DDoS intrusion. The method has good performance and can timely and accurately detect capacity exhaustion DDoS intrusion and apply mitigation. In addition, the invention also effectively reduces system overhead and achieves lower switch resource consumption.

[0008] The technical solution adopted by the present invention to achieve the above-mentioned objectives is: the capacity exhaustion type DDoS intrusion detection and mitigation method mainly includes five steps: network information collection and state quantification, preliminary judgment of network state, online time series analysis, final judgment of network state and application of mitigation measures.

[0009] 1. Network information collection and status quantification: Using the time window as the basic unit of network information collection, the Sketch operation is dynamically activated based on the traffic information of the previous time window, and network traffic information is collected through Sketch: number of packets, source IP cardinality, and based on this information, the information entropy calculation method in the data plane is used, and the calculated information entropy is used as a quantitative indicator of the network status.

[0010] 2. Preliminary judgment of network status. Make a preliminary judgment of network status based on the entropy value calculated in the current time window, and set two thresholds: θ1 is the easy-to-distinguish attack judgment threshold, and θ2 is the complex judgment threshold for suspicious status. Threshold θ1 is used to distinguish high-probability intrusion flows from non-high-probability intrusion flows, and threshold θ2 is used to distinguish suspicious intrusion flows from normal flows. If the value exceeds θ1, go directly to step 5; if the value is less than θ1 and greater than θ2, go to step 3; if the value is less than θ2, it is judged as a normal flow and no further analysis or operation is performed.

[0011] 3. Online time series analysis. The controller polls each target IP address dst A time series of fixed length is maintained, which is the network state information entropy of the sequential time window. The sequence information is used to make further complex judgments on suspicious flows. The controller uses a time series analysis method based on chaos theory. First, the time series data is reconstructed in phase space, and then the detrended fluctuation analysis method is used to calculate the fractal cone number of the maintained time series.

[0012] 4. Final judgment of network status. Combine the preliminary judgment result of step 2 and the fractal cone number calculated according to step 3 to make the final judgment of the network status. If step 2 determines that the current network status is only suspicious, then go to step 3 to analyze and calculate the fractal cone number of the time series corresponding to the network status in the current time window. When the fractal cone number is less than 0.5, it is judged that an intrusion event has occurred in the network, and then go to step 5. If it is greater than or equal to 0.5, it is judged that no intrusion event has occurred, and no further operation is applied.

[0013] 5. Apply mitigation measures: The control plane reads the network topology and models the network as a directed graph. If an intrusion is detected, IP backtracking is performed. At the same time, filtering flow table rules are added to the data plane to specifically filter the attack source IP obtained through backtracking. Beneficial Effects

[0014] The capacity exhaustion type DDoS intrusion detection and mitigation method proposed in the present invention collects network status information by setting a dynamic Sketch call module in a programmable switch at the edge of the network, performs two-stage timing analysis for detection through the network status characteristics provided by the network status information, and performs graph modeling by reading the network topology to perform attack source IP backtracking, and finally adds filtering flow table rules for the attack source IP to complete the mitigation. The method can realize efficient and high resource utilization capacity exhaustion type DDoS intrusion detection and achieve rapid mitigation. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of dynamic Sketch call in the data plane in the method of the present invention.

[0016] Figure 2 The figure is a schematic diagram of two-stage intrusion detection judgment in the control plane and the data plane in the method of the present invention.

[0017] Figure 3 The schematic diagram is a diagram of IP backtracking and mitigation based on network topology modeling in the method of the present invention.

[0018] Figure 4 The flowchart of the capacity exhaustion type DDoS intrusion detection and mitigation method based on online timing analysis. DETAILED DESCRIPTION

[0019] The present invention is further described below in conjunction with the accompanying drawings.

[0020] Figure 1 This is a schematic diagram of the dynamic Sketch call in the data plane in the present invention. The present invention uses Sketch technology to measure the network status. At the same time, in order to solve the shortage of resources in the programmable switch, a functional module for dynamically calling Sketch to measure and estimate the network status is established. In this functional module based on Sketch technology, multiple dynamically activated Count-Min Sketches and a fixed HyperLogLog Sketch are used as basic units. The specific measurement of the network status is based on a time window of fixed length as an observation measurement interval. In the i-th time window TW i In the process, the dynamic activation amount is set based on the network status information provided by the previous time window, so as to dynamically activate and call Sketch. The dynamic activation amount is calculated as: Δ i =Entropy i-1 -θ chaos At the same time, according to Δ i Perform range matching in the data plane and set a range node set r of size N for specific situations.i (i=0,…N-1), where r0=0, if Δ i ∈[r i-1 ,r i ), then i Count-Min Sketches are activated for measurement and estimation.

[0021] Figure 2 The schematic diagram of the two-stage intrusion detection judgment in the data plane and the control plane in the method of the present invention. First, a preliminary judgment is made in the data plane whether an intrusion has occurred, and the current time window TW collected by Sketch is used to determine whether an intrusion has occurred. i (i∈(1,....n)) inner package number Num packet Source IP base number Num srcIP Calculate the entropy in the data plane and use the joint entropy calculation to calculate the time window TW i Entropy of internal network status i , the specific calculation formula is as follows: Among them, p(x,y) is in the time window TW i The joint probability that the number of packets is x and the source IP cardinality is y. Here, the statistical frequency is used to reflect the joint probability of the two measurement features.

[0022] In the completion time window TW i Entropy i After the calculation, a preliminary judgment is made in the data plane, and two thresholds are set: one is the easy-to-distinguish attack judgment threshold θ1, and the other is the suspicious state complex judgment threshold θ2. i Perform range matching. If Entropy i <θ1, it is considered normal flow. If θ1≤Entropy i ≤θ2, then it is considered a suspicious flow. If Entropy i >θ2, it is judged as an abnormal flow. If it is judged as a normal flow, no operation is performed; if it is judged as an abnormal flow, digest is directly used to upload the time information of the abnormal time window and the specific target IP address to the control plane, and the control plane takes specific mitigation measures; if it is judged as a suspicious flow, the suspicious state judgment mark, the time information of the suspicious time window and the specific target IP address are uploaded, and the control plane makes further judgments.

[0023] When a suspicious flow is judged and relevant information is uploaded, the control plane performs suspicious state analysis and judgment based on chaos theory, and finally makes a final judgment. The control plane maintains a time series of fixed length m by polling every τ time. The value of this time series is the network state measurement value Entropy of each time window. Let this sequence be x(t). This polling method directly reconstructs the phase space of this sequence: X(t)=(x(t),x(t+τ),x(t+2τ),…,x(t+(m-1)τ)) Wherein, t=1, 2, ..., N-(m-1), and τ is the delay time.

[0024] Phase space reconstruction usually requires determining the delay time τ and the embedding dimension m, that is, determining the polling time interval of the control plane and maintaining a fixed length m of the time series. First, the delay time τ is determined by the autocorrelation function method. This method uses the autocorrelation function R(τ) to describe the similarity between the time series and itself after time τ. The τ that makes R(τ) drop to zero or close to zero for the first time is selected as the delay time for phase space reconstruction. The formula of the autocorrelation function is as follows: in, is the average value of the time series, and N is the length of the time series. Secondly, the embedding dimension d is determined by the Cao method, which is an improvement of the false nearest point method. The Cao method improves the false nearest point method by defining two functions E1(d) and E2(d) to determine the embedding dimension of the time series. By calculating and observing when E1(d) begins to approach 1 and whether E2(d) is stable at a constant, the d that meets these conditions is the minimum embedding dimension, that is, the optimal embedding dimension. The calculation formulas of the two are: Where N is the length of the time series. i is the i-th point in the sequence. τ is the time delay, d is the embedding dimension. j(i,d) is the index of the nearest neighbor of the i-th point when the embedding dimension is d.

[0025] After determining the basic composition of the time series after phase space reconstruction, the detrended fluctuation analysis (DFA) method is used for further analysis. In the phase space, the cumulative deviation function Y for each dimension j = 1, 2, 3, ..., m needs to be calculated first. j (i): Construct a cumulative deviation matrix from the cumulative deviation function for each dimension: Y(i)=[Y1(i),Y2(i),…,Y m(i)] The reconstructed time series is divided into non-overlapping windows of length s, and polynomial fitting is performed in each window to remove the trend. Assume that the sequence length is N and the number of windows is For each window v=1,2,…,N s , calculate the cumulative deviation after detrending: In each window v, after removing the local trend, the volatility function is calculated Finally, calculate the overall fluctuation function F(s): By plotting log(F(s)) versus log(s), we can obtain the scaling exponent, or fractal cone number α, to reveal the long-range correlation of the time series: F(s)~s α

[0026] When α<0.5, it means that the time series has anti-persistence, that is, the value of the time series tends to be opposite to the previous value. At this time, it can be judged that an intrusion state has occurred, and the mitigation mechanism is triggered; when α≥0.5, it means that the time series has persistence, indicating that the phase space constructed by the suspicious state maintains the normal state, so it is judged that no intrusion state has occurred and the mitigation mechanism is not triggered.

[0027] Figure 3 Schematic diagram of IP backtracking and mitigation based on network topology modeling in the method of the present invention. The method identifies the IP address according to the target IP address uploaded when the intrusion is detected. dst The network topology information is read and the network is modeled as a directed graph G = (V, E), where V represents the switch node and E represents the network link. A set of flow table forwarding rules [T0, ..., TM] associated with each individual node are set as different edges. A reverse depth-first search algorithm is used to find a path as quickly as possible. Specifically, starting from the target node, a reverse depth-first search is used to find all edges pointing to it, and the visited nodes are recorded to avoid repeated visits. The predecessor node is continued to be recursively searched until there is no predecessor node (that is, a source node is found).

[0028] After finding the source node, that is, the source IP address of the intrusion, the controller first establishes a connection with the P4 switch through the P4Runtime API and creates a session. Then, it obtains the P4Info description file. The P4Info file describes the structure of the P4 program, including tables, actions, data packet formats and other information. The controller obtains and loads this file to understand the flow table structure and definition. After that, the controller creates a table entry according to the rules that need to be inserted. For example, if the source IP address of the intrusion is 192.168.0.1, then the match field is set to the source IP address 192.168.0.1, and the action is drop. Next, the controller calls the WriteTableEntry or ModifyTableEntry method to send the constructed table entry to the P4 switch. After receiving the table entry, the switch verifies the legitimacy of the entry and performs the corresponding action. The controller can query the switch status to confirm that the rules have been correctly issued and effective.

[0029] Figure 4 The flowchart of the capacity exhaustion type DDoS intrusion detection and mitigation method based on online time series analysis. The capacity exhaustion type DDoS intrusion detection and mitigation method mainly includes five parts: network information collection and state quantification, preliminary judgment of network state, online time series analysis, final judgment of network state and application of mitigation measures. For each network edge target IP address, in the data plane, based on the last time window information, the module called by dynamic Sketch is used to collect network state information. The module uses Count-Min Sketch to collect the number of data packets for the target IP address, and uses HyperLogLog Sketch to collect the source IP cardinality corresponding to the data packets for the target IP address, so as to measure and estimate the network state. At the same time, using these two network status features, the information entropy calculation form is adopted in the data plane, and the information entropy is used as a comprehensive reflection of the network status. The information entropy calculation result is compared with the easy-to-distinguish attack judgment threshold θ1 and the complex judgment threshold θ2 of the suspicious state in the data plane to make a preliminary judgment. If it exceeds θ1, the mitigation mechanism is directly triggered through digest. If it does not exceed θ1 but exceeds θ2, the control plane is triggered to further analyze and judge the suspicious state; in the control plane, for each target IP address, a time window is used as a division, and a time series with a fixed length value of the information entropy in the time window is maintained. The final judgment of the suspicious state is made through the phase space reconstruction and detrended fluctuation analysis method based on chaos theory. If it is judged as an intrusion, the mitigation mechanism is directly triggered in the control plane; in terms of mitigation, the control plane reads the network topology, models the graph, and uses the flow table rules as the edges of the graph to reversely trace the source, find the source IP address of the intrusion, and insert the flow table rules for filtering the source IP address to achieve intrusion mitigation.

Claims

1. Volumetric DDoS intrusion detection and mitigation method based on online timing analysis, It is characterized in that In the data plane, Sketch is used to estimate the eigenvalues ​​and calculate the information entropy of the observation time window. Based on this, the first-stage preliminary judgment is made to complete the filtering detection of intrusion situations that are easy to distinguish. In the control plane, chaos theory is used to perform online time series analysis, based on which the second-stage final judgment is made to complete the screening detection of intrusion situations that are more difficult to distinguish. After the detection is completed, the control plane performs IP source tracing and inserts flow table rules to mitigate the intrusion into the data plane. The capacity exhaustion type DDoS intrusion detection and mitigation method based on online timing analysis includes the following steps: Step 1: Network information collection and status quantification: Using the time window as the basic unit for network information collection, dynamically activate Sketch based on the traffic information of the previous time window, and collect network traffic information through Sketch: number of packets, source IP cardinality, and use the method of calculating information entropy in the data plane based on this information, and use the calculated information entropy as a quantitative indicator of the network status; Step 2: Preliminary judgment of network status: Make a preliminary judgment of the network status based on the entropy value calculated in the current time window, and set two thresholds, θ 1 is the easy-to-distinguish attack judgment threshold, θ 2 is the complex judgment threshold of suspicious state, threshold θ 1 It is used to distinguish high-probability intrusion flows from non-high-probability intrusion flows, and the threshold θ 2 Used to distinguish suspicious intrusion flows from normal flows, exceeding θ 1 Go directly to step 5, less than θ 1 Greater than θ 2 Go to step 3, less than θ 2 It is determined to be a normal flow and no further analysis or operation is performed; Step 3: Online time series analysis: The controller polls each target IP address dst A fixed-length time series is maintained, which is the network state information entropy of the sequential time window. The sequence information is used to make further complex judgments on suspicious flows. The controller uses a time series analysis method based on chaos theory. First, the phase space of the time series data is reconstructed, and then the fractal cone number of the maintained time series is calculated using the detrended fluctuation analysis method. Step 4, final judgment of network status: Combine the preliminary judgment result of step 2 and the fractal cone number calculated according to step 3 to make the final judgment of the network status. If step 2 determines that the current network status is only suspicious, then go to step 3 to analyze and calculate the fractal cone number of the time series corresponding to the network status in the current time window. When the fractal cone number is less than 0.5, it is judged that an intrusion event has occurred in the network, and then go to step 5. If it is greater than or equal to 0.5, it is judged that no intrusion event has occurred, and no further operation is applied. Step 5: Apply mitigation measures: The control plane reads the network topology and models the network into a directed graph. If an intrusion is detected, IP backtracking is performed. At the same time, filtering flow table rules are added to the data plane to specifically filter the attack source IP obtained through backtracking.

2. According to the volumetric exhaustion type DDoS intrusion detection and mitigation method described in claim 1, It is characterized in that In step 1, the information of the previous time window is used to dynamically activate and use multiple Sketches to complete the collection of network information. At the same time, information entropy is used as a quantification of the network status. The above operations and calculations are completely performed in the data plane. Sketch technology is a set of probabilistic data structures used to process large-scale data streams and data sets. This technology can quickly estimate various statistics with limited memory space and allowed errors when the data stream cannot be fully stored. The detection method uses Count-Min Sketch and HyperLogLog Sketch. Count-Min Sketch is a two-dimensional array based on hash functions, which is used to maintain frequency counts in data streams through multi-hash function mapping, so as to quickly estimate element frequencies while allowing errors. HyperLogLog Sketch is a probabilistic data structure based on hash functions and bitmaps, which can effectively estimate the cardinality of large-scale data sets by recording the maximum tail zero length of hash values. In the data plane, multiple basic Sketch units are dynamically activated by maintaining and keeping a value Δ, where the value Δ is the information entropy and the chaos threshold θ of the current time window. chaos According to the network conditions, the difference between Δ and Δ is adaptively matched to dynamically activate the basic Sketch unit, thereby completing the estimation and statistics of the number of data packets and the source IP cardinality; The basic Sketch unit that can be dynamically activated is the Count-Min Sketch, which is used to estimate the number of data packets based on the network status; at the same time, a HyperLogLog Sketch is fixedly set to estimate the source IP cardinality based on the network status.

3. According to the volumetric exhaustion type DDoS intrusion detection and mitigation method described in claim 1, It is characterized in that The complex judgment method of suspicious status in step 3 is performed online on the control plane, for each target IP address IP dst A fixed-length time series is maintained, and the second-stage judgment is carried out using a time series analysis method based on chaos theory. This method can be divided into two steps: Step 3.1: First, perform phase space reconstruction. Use different time delays of the given one-dimensional time series {x(i)}(i∈[1,n]) to construct an m-dimensional phase space vector, that is, construct the following form: X(t)=(x(t),x(t+τ),x(t+2t),…,x(t+(d-1)τ)) Wherein, t=1, 2, ..., N-(d-1), τ is the delay time; Step 3.2, then use the detrended fluctuation analysis method to complete the judgment of the suspicious state, and calculate the fluctuation function through the following formula: Among them, s is the window size, that is, the time scale, is the number of windows, N is the length of the time series; is the cumulative deviation function, where x(k) in Y(i) is a time series data point, is the mean of the time series; is the local trend fitted using the least squares method within the window v; Through detrended fluctuation analysis, the following power law relationship is obtained: F(s)~s α Among them, α is the scaling index. α<0.5 indicates that the time series has anti-persistence, that is, the value of the time series tends to be opposite to the previous value. At this time, it can be judged that an intrusion state has occurred. When α≥0.5, it means that the time series has persistence, indicating that the phase space constructed by the suspicious state maintains the normal state, so it is judged that no intrusion state has occurred.

4. According to the volumetric exhaustion type DDoS intrusion detection and mitigation method described in claim 1, It is characterized in that Step 4 is the final judgment of the network traffic status in the current time window jointly decided by the control plane and the data plane. In the preliminary judgment stage of step 2, if an intrusion is directly detected in the data plane, the programmable switch uses digest to report warning information and trigger the mitigation mechanism in step 5; in the online timing analysis stage of step 3, the inspection result of the intrusion is obtained through analysis, and the mitigation mechanism in step 5 is directly triggered in the controller.

5. According to the volumetric exhaustion type DDoS intrusion detection and mitigation method described in claim 1, It is characterized in that Step 5 applies mitigation measures based on the final determined attack state, and models the network as a directed graph G = (V, E), where V represents the switch node, E represents the network link, and a set of flow table forwarding rules [T0, ..., TM] associated with each individual node are set as different edges. When abnormal traffic is detected, the abnormal time range, location, matching field and other information are input. Starting from the location where the abnormality is detected, that is, the end host, reverse trace along the flow table rules to find all the inlet switch nodes that may cause the abnormal traffic, reversely trace the attacker's source IP based on these inlets, and insert filtering flow table rules into the data plane for these attack source IPs.