Graph generation method, graph generation device, and program

The graph generation method addresses the limitations of existing graphlet analysis by creating a communication behavior graph that accounts for time-based changes in host interactions, resulting in improved network monitoring and analysis capabilities.

WO2024241543A1PCT designated stage expired Publication Date: 2024-11-28NT T INC
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
PCT/JP2023/019382
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-24
Publication Date
2024-11-28

AI Technical Summary

Technical Problem

Existing graphlet analysis methods fail to account for behavioral changes over time, leading to inaccurate host profiles and an incomplete understanding of communication behavior patterns in network monitoring and analysis.

Method used

A graph generation method that divides traffic data into time slots, generates graphlets for each slot, defines communication behaviors as nodes, and transitions between behaviors as edges, creating a communication behavior graph to capture the dynamic nature of host interactions.

Benefits of technology

This approach provides a more accurate and comprehensive visualization of communication behavior patterns, enabling detailed network monitoring and analysis by capturing the evolution of host characteristics over time.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure JP2023019382_28112024_PF_FP_ABST
    Figure JP2023019382_28112024_PF_FP_ABST
Patent Text Reader

Abstract

This graph generation method, which is executed by a graph generation device, includes: an acquisition step for acquiring traffic data pertaining to a host being analyzed; and a graph generation step for dividing the traffic data for each time slot, generating a graphlet of each time slot from the divided traffic data of each time slot, and generating a communication behavior graph in which a communication behavior corresponding to each graphlet serves as a node and in which a transition between communication behaviors serves as an edge.
Need to check novelty before this filing date? Find Prior Art

Description

Graph generation method, graph generation device, and program

[0001] The present invention relates to a technique for analyzing the behavioral patterns of communication terminals (hosts) on a network.

[0002] Host profiles, which identify the characteristics of hosts on a network, are important for network monitoring and analysis. Host profiles estimate attribute information that indicates the main characteristics of hosts on a network.

[0003] Specific examples of host attribute information include attribute information indicating client / server functions such as "Web / mail / NTP / DNS," and attribute information indicating malware communication behavior such as "SCAN communication / flooding attack."

[0004] As a conventional technique, a host profile based on graphlet analysis has been proposed (Non-Patent Document 1).

[0005] Himura, Yosuke, et al. "Synoptic graphlet: Bridging the gap between supervised and unsupervised profiling of host-level network traffic." IEEE / ACM Transactions on Networking 21.4 (2012): 1284-1297.

[0006] Generally, a communication host runs multiple different programs and processes at any given time. By capturing a series of communication behaviors of a communication host as a communication behavior pattern, it becomes possible to analyze the host's behavior in more detail. In other words, to create an accurate host profile, it is necessary to analyze the order in which multiple communication behaviors are executed.

[0007] However, the existing graphlet analysis method disclosed in Non-Patent Document 1 does not take into account behavioral changes over time, and therefore cannot investigate communication behavior patterns of hosts.

[0008] The present invention has been made in view of the above points, and has an object to provide a technique for generating a communication behavior graph that indicates the communication behavior patterns of a host.

[0009] According to the disclosed technology, there is provided a graph generation method executed by a graph generation device, the graph generation method comprising: an acquisition step of acquiring traffic data of a host to be analyzed; and a graph generation step of dividing the traffic data into time slots, generating graphlets for each time slot from the divided traffic data for each time slot, and generating a communication behavior graph in which communication behaviors corresponding to the graphlets are nodes and transitions between communication behaviors are edges.

[0010] According to the disclosed techniques, techniques are provided for generating a communication behavior graph that indicates communication behavior patterns of hosts.

[0011] 6 is a diagram illustrating an example of a graphlet. FIG. 7 is a diagram illustrating an example of a graphlet. FIG. 8 is a diagram illustrating an example of the overall configuration of a system in an embodiment of the present invention. FIG. 9 is a diagram illustrating an example of the functional configuration of the graph generation device 100. FIG. 10 is a diagram illustrating an example of a line graph for each packet. FIG. 11 is a diagram illustrating an example of a graphlet generated from the line graph of FIG. 6. FIG. 12 is a flowchart illustrating the operation of the graph generation device 100. FIG. 13 is a diagram illustrating an edge generation method. FIG. 14 is a diagram illustrating an edge generation method. FIG. 15 is a diagram illustrating an example of a communication behavior graph. FIG. 16 is a diagram illustrating an example of the hardware configuration of the graph generation device 100.

[0012] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. The embodiment described below is merely an example, and the embodiment to which the present invention is applied is not limited to the following embodiment.

[0013] In the following, first, the problems of the conventional technology and the technology according to the present embodiment will be described in detail, and then the technology according to the present embodiment will be described. Note that the explanation of the problems below is not publicly known.

[0014] (Regarding Prior Art) As mentioned above, a host profile using a graphlet analysis method has been proposed as prior art (Non-Patent Document 1). The graphlet analysis method is a technique for visualizing host behavior as a graph and improving interpretability.

[0015] A graphlet is composed of multiple node strings, each of which represents a flow attribute (source IP address, source port number, destination IP address, destination port number, protocol number). A node string has one or more nodes, and each node represents the value of its attribute.

[0016] Graphlets are created using traffic information divided by host. Specifically, a certain number of packets (hundreds to thousands) are collected from each host and visualized as a graphlet. Generation of identical communications is omitted, and new nodes and edges are generated only when different communication patterns occur.

[0017] 1 and 2 show examples of graphlets for a certain host. From the graphlet example shown in Fig. 1, it can be seen that a certain host with one source IP address is communicating with multiple destination IP addresses and multiple destination ports with source port number 80.

[0018] Since port number 80 is used for HTTP communication and communication is being made to multiple destination IP addresses, it is clear that this host is highly likely to be a Web server.

[0019] The graphlet example shown in Figure 2 indicates that communication is being sent from a single source port to multiple destination ports for a single destination IP address. This suggests that the source host is likely launching a port scan attack against the destination host.

[0020] In this way, host profiling becomes possible by creating graphlets from traffic information.

[0021] (About the Issues) Conventional host profiles using graphlet analysis assign a single attribute to a host by analyzing its communication behavior from communication traffic information collected over a short period of time.

[0022] However, in general, communication hosts exhibit different communication behaviors at different times. Therefore, with existing methods, the same host may be assigned different attributes depending on the period of communication traffic collection, which creates issues with the accuracy of host profiles. Furthermore, simply identifying a single attribute, as with existing methods, does not allow for a complete picture of a host's communication behavior.

[0023] (Outline of the embodiment) In this embodiment, in order to solve the above-mentioned problem, the existing graphlet analysis method is extended in the time direction, thereby making it possible to analyze the transition of the communication behavior of a host (communication behavior pattern).

[0024] In this embodiment, a graph generation device 100 (described later) represents the above communication behavior patterns using a graph in which communication behaviors are nodes and transitions of communication behaviors over time are edges. The graph generation device 100 determines the communication behavior corresponding to the traffic to be analyzed using a basic graphlet constructed from a single communication behavior traffic.

[0025] (System Configuration Example) Fig. 3 shows an example of the overall configuration of a system according to this embodiment. As shown in Fig. 3, in this system, a communication terminal 300 is connected to the Internet 500 via an ISP (Internet Service Provider) network 400. In addition, a graph generation device 100 is connected to a traffic capture server 200, and the traffic capture server 200 is connected to the ISP network 400.

[0026] The traffic capture server 200 acquires (captures) traffic data of communications by the communication terminal 300 from the ISP network 400. The captured traffic data is passed to the graph generation device 100. The graph generation device 100 uses the traffic data to generate and output a communication behavior graph.

[0027] 4 shows an example of the functional configuration of the graph generation device 100. As shown in FIG. 4, the graph generation device 100 includes an acquisition unit 110, a graph generation unit 120, an output unit 130, and a storage unit 140.

[0028] The acquiring unit 110 acquires traffic data from the traffic capture server 200 and stores the acquired traffic data in the storage unit 140. The acquiring unit 110 may also include the functionality of a traffic capture server. In this case, the traffic capture server 200 is not required, and the acquiring unit 110 can acquire traffic data from the ISP network 400.

[0029] The graph generation unit 120 reads traffic data from the storage unit 140 and generates a communication behavior graph using the traffic data. The process for generating a communication behavior graph will be described later. The output unit 130 outputs the communication behavior graph generated by the graph generation unit 120. For example, the output unit 130 visualizes the communication behavior graph and displays it on a display.

[0030] (Graphlet terminology definition) Here, the term graphlet will be defined. When five general traffic attribute information (source IP address, destination IP address, source port number, destination port number, protocol number) is used as input data (traffic data) of a graphlet, each attribute information is sequentially assigned a column number, and the graphlet is then sorted as follows: 1 , ..., A 5 A graphlet is expressed as 1 From A 5 The columns are arranged in order from left to right.

[0031] n i is A i It represents the number of nodes in the column. i Column n i Nodes are v 1,i ,... ,v n_i,i Here, in this specification, "v n_i,i ", the subscript in the subscript is expressed using "_". n_i,iIn the example, "n_i" is "n i " is the intention.

[0032] i:j is A i From A j represents the direction to, where j = i±1.

[0033] o i:j is A i From A j represents the number of nodes that have only one edge.

[0034] u i:j is A i From A j A for the edge in the direction i represents the average degree of the nodes in

[0035] Node V k,i The in-degree to (i-1: number of edges in the i direction) is d k,i:i-1 and node v k,i The out-degree (number of edges in the i:i+1 direction) of k,i:i+1 It is expressed as:

[0036] α i:j is A i From A j The node with this maximum degree is called v h_i,i Let's say.

[0037] β i:i+1 = d h_i,i:i-1 and this is A i From A j The node v with the maximum degree in the direction h_i,i is the in-degree to

[0038] (Graphlet Feature Vector) Using the above definition, the five types of graphlet features (n i , o i:j , u i:j , α i:j , β i:j ) constitutes a graphlet feature vector x.

[0039] For example, A 1 , ..., A 5 When a graphlet is composed of five attribute information, n i is n1 ,...,n 5 It becomes the fifth dimension, and o i:j Hao 1:2 , o 2:1 , ..., o 5:4 8 dimensions, and u i:j Ha 1:2 , u 2:1 , ..., u 5:4 8 dimensions, and α i:j is α 1:2 , α 2:1 , ..., α 5:4 8 dimensions, and β i:j is β 2:1 , β 2:3 , ..., β 4:5 The feature vector x has a total of 35 dimensions.

[0040] Graphlet g i , g j The distance γ (g i , g j ) is the Euclidean distance Σ between each feature vector x and y. n i=1 (x i -y i ) 2 (n is the number of dimensions of the feature vectors x and y).

[0041] (Basic Graphlet Construction Method) Next, a basic graphlet construction method executed by the graph generation unit 110 will be described.

[0042] Traffic data D is a packet p i When traffic data D is composed of N packets, D=(p 1 , ..., p N ) An example of the traffic data D is shown in FIG.

[0043] The graph generating unit 120 reads the traffic data from the storage unit 140 and generates a line graph for each packet. Fig. 6 shows an example of a line graph for each packet constructed from the traffic data D in Fig. 5 .

[0044] The graph generation unit 120 generates (constructs) a graphlet by calculating the union of line graphs and deleting overlapping nodes and edges. Fig. 7 shows an example of a graphlet generated from the line graph of Fig. 6.

[0045] As a preliminary preparation for generating the communication behavior graph of this embodiment, the graph generation unit 120 generates a basic graphlet in advance from traffic data of a single communication behavior. A single communication behavior label is assigned to the basic graphlet. Note that the graphlets shown in Figures 1 and 2 are examples of basic graphlets that can read only a single communication behavior.

[0046] Basic graphlet G i Traffic data D constituting i The number of packets in |D i | and the communication behavior label is l i The basic graphlets are generated for each communication behavior, and the set of basic graphlets is expressed as B = (G 1 , ..., G N_base ) is expressed as N base represents the number of graphlets included in the basic graphlet set.

[0047] The graph generating unit 120 also generates feature vectors for all the basic graphlets. The set of generated basic graphlets and the feature vectors for each basic graphlet are stored in the storage unit 140.

[0048] (Processing for Generating Communication Behavior Graph) Next, a procedure (generation algorithm) for generating a communication behavior graph executed by the graph generation device 100 will be described with reference to FIG.

[0049] <S101> In S101, the acquisition unit 110 collects traffic data of the host to be analyzed and stores the collected traffic data in the storage unit 140. Each piece of traffic data is stored in the storage unit 140 in association with a time.

[0050] The graph generation unit 120 reads the traffic data of the host to be analyzed from the storage unit 140 and divides it into predetermined time slots. The length of the time slots varies depending on the use case, but since one communication behavior is assigned to each time slot, a few seconds to a few minutes is appropriate. However, this is not limited to a few seconds to a few minutes.

[0051] <S102> In S102, the graph generation unit 120 generates a graphlet G from the traffic data at the divided time t (time slot t). t is generated and the feature vector of the graphlet is calculated.

[0052] <S103> In S103, the graph generation unit 120 generates a graphlet G t The distance between the graphlet and all the basic graphlets is calculated, and the communication behavior label of the basic graphlet with the smallest distance is determined as the communication behavior for the time slot. More details are as follows.

[0053] The graph generation unit 120 calculates the communication behavior label l in the time slot t by the following formula: t Determine.

[0054] l t =l argmin_i (γ(G t , G i )) where argmin represents the input argument that gives the minimum value of the function, and γ(G t , G i ) is graphlet G t and the basic graphlet G i represents the Euclidean distance between the feature vectors. i is the basic graphlet G i This formula represents the communication behavior label of the graphlet G t and the basic graphlet set B = (G 1 , ..., G N_base ) and assigns the communication behavior label of the basic graphlet with the smallest distance as the communication behavior for the time slot.

[0055] In the desired communication behavior graph G = (V, E), the node set V has a communication behavior label l t E is a set of edges.

[0056] <S104> The graph generation unit 120 performs the processes of S102 and S103 on all the divided traffic data. When the processes on all the divided traffic data are completed, the process proceeds to S105.

[0057] <S105> In S105, the graph generation unit 120 adds edges representing transitions between communication behaviors between nodes based on the time information and edge conditions of each node. This generates a communication behavior graph G = (V, E). The output unit 130 outputs the communication behavior graph G = (V, E). Note that the generated communication behavior graph G = (V, E) may be stored in the storage unit 140 without being output.

[0058] More specifically, the graph generation unit 120 generates edges in the following steps (1) and (2): In the following steps, a and bεN are parameters of this algorithm.

[0059] (1) Edge Generation The graph generation unit 120 generates edges for all communication actions from the communication action of time slot t to time slot t+a for each time slot t.

[0060] (2) Edge Determination Next, if there are less than b overlapping edges between nodes, the graph generation unit 120 determines that the transition in communication behavior occurred accidentally and deletes all edges between the nodes. On the other hand, if there are b or more overlapping edges between nodes, it determines that the transition in communication behavior was inevitable and leaves only one edge between the nodes.

[0061] A specific example of steps (1) and (2) will be described with reference to Fig. 9. In the example of Fig. 9, the communication behavior at t = 1, 2, 3 is 1 and the communication behavior at t = 4, 5, 6 is l 3 and the communication behavior at t=7 is l 4 and the communication behavior at t=8, 9 is l 2In FIG. 9, one communication behavior is represented as one node.

[0062] Here, a = 2. First, the communication action l at t = 1 1 From t = 1 + 2 = 3, all communication actions (communication action l at t = 2) 1 , communication behavior l at t=3 1 ) but with the same communication behavior l 1 So no edges are generated.

[0063] Next, communication behavior l at t=2 1 From all communication actions up to t = 4 (communication action l at t = 3) 1 , communication behavior l at t=4 3 ) by creating an edge to 1 Communication behavior from the node 3 By performing similar edge generation up to t=9, edges as shown in FIG.

[0064] Here, b = 2. By procedure (2), the number of overlapping edges is less than 2. 3 Communication behavior from the node 2 The edge to the node " is deleted, and further, the edge between the nodes is reduced to one, thereby obtaining the communication behavior graph shown in Fig. 10. Fig. 11 shows a more detailed example of the communication behavior graph.

[0065] (Hardware Configuration Example) The graph generation device 100 described in this embodiment can be realized, for example, by causing a computer to execute a program. This computer may be a physical computer or a virtual machine on the cloud.

[0066] That is, the graph generation device 100 can be realized by using hardware resources such as a CPU and memory built into a computer to execute a program corresponding to the processing performed by the graph generation device 100. The program can be recorded on a computer-readable recording medium (such as a portable memory) and can be saved or distributed. The program can also be provided via a network such as the Internet or email.

[0067] Fig. 12 is a diagram showing an example of the hardware configuration of the computer. The computer in Fig. 12 includes a drive device 1000, an auxiliary storage device 1002, a memory device 1003, a CPU 1004, an interface device 1005, a display device 1006, an input device 1007, an output device 1008, and the like, all of which are interconnected by a bus BS. The computer may further include a GPU.

[0068] The program that realizes the processing on the computer is provided by a recording medium 1001, such as a CD-ROM or a memory card. When the recording medium 1001 storing the program is set in the drive device 1000, the program is installed from the recording medium 1001 to the auxiliary storage device 1002 via the drive device 1000. However, the program does not necessarily have to be installed from the recording medium 1001, but may be downloaded from another computer via a network. The auxiliary storage device 1002 stores the installed program as well as necessary files, data, etc.

[0069] The memory device 1003 reads and stores the program from the auxiliary storage device 1002 when an instruction to start the program is received. The CPU 1004 realizes functions related to the graph generation device 100 in accordance with the program stored in the memory device 1003. The interface device 1005 is used as an interface for connecting to a network, etc. The display device 1006 displays a GUI (Graphical User Interface) or the like according to the program. The input device 1007 is composed of a keyboard, mouse, buttons, a touch panel, etc., and is used to input various operation instructions. The output device 1008 outputs the results of calculations.

[0070] (Summary and Effects of the Embodiment) As explained above, in the technology explained in this embodiment, in response to the problem of the need to analyze communication behavior patterns for host profiles, individual communication behaviors are examined using graphlet analysis, and a communication behavior graph is constructed in which communication behaviors are nodes and transitions between communication behaviors are edges. The communication behavior graph makes it possible to grasp the characteristics of hosts in more detail, and is therefore useful in monitoring and analyzing networks. More specifically, the technology according to this embodiment provides, for example, the following effects (1) and (2).

[0071] (1) The overall picture of communication behavior can be visualized.

[0072] The communication behavior graph allows you to see the entire communication behavior at a glance, making it possible to grasp the overall picture of the communication situation. This makes it possible to discover characteristics of hosts that could not be identified by analyzing individual communication behavior.

[0073] (2) It becomes possible to discover problems through structural analysis of belief behavior.

[0074] By using a communication behavior graph, structural analysis of graph theory makes it possible to discover problems based on the connectivity and distribution between communication behavior nodes. For example, in a communication behavior graph, the importance of nodes can be measured, making it possible to identify the most important communication behaviors.

[0075] The following additional notes are provided regarding the above-described embodiments.

[0076] <Additional Notes> (Additional Item 1) A graph generation method executed by a graph generation device, comprising: an acquisition step of acquiring traffic data of a host to be analyzed; and a graph generation step of dividing the traffic data by time slots, generating graphlets for each time slot from the divided traffic data for each time slot, and generating a communication behavior graph in which communication behaviors corresponding to the graphlets are nodes and transitions between communication behaviors are edges. (Additional Item 2) The graph generation method according to Additional Item 1, wherein in the graph generation step, the graph generation device determines a communication behavior corresponding to the graphlet based on the distance between the graphlet and a basic graphlet to which a communication behavior is assigned. (Additional Item 3) The graph generation method according to Additional Item 2, wherein in the graph generation step, the graph generation device determines the communication behavior of a basic graphlet that has the smallest distance from the graphlet among a set of basic graphlets as the communication behavior of the graphlet. (Supplementary Item 4) The graph generation method according to any one of Supplementary Items 1 to 3, wherein in the graph generation step, the graph generation device generates edges for all communication behaviors from a communication behavior in a certain time slot to a time slot that is a predetermined time after the certain time slot, and generates the communication behavior graph by deleting all edges between the nodes if the number of overlapping edges between nodes is less than a predetermined number. (Supplementary Item 5) A graph generation device including: a memory; and at least one processor connected to the memory, wherein the processor acquires traffic data of an analysis target host, divides the traffic data by time slots, generates graphlets for each time slot from the divided traffic data for each time slot, and generates a communication behavior graph in which communication behaviors corresponding to the graphlets are nodes and transitions between communication behaviors are edges. (Supplementary Item 6) A non-transitory storage medium storing a program for causing a computer to execute each step of the graph generation method according to any one of Supplementary Items 1 to 4.

[0077] Although the present embodiment has been described above, the present invention is not limited to such a specific embodiment, and various modifications and changes are possible within the scope of the gist of the present invention described in the claims.

[0078] REFERENCE SIGNS LIST 100 Graph generation device 110 Acquisition unit 120 Graph generation unit 130 Output unit 140 Storage unit 200 Traffic capture server 300 Communication terminal 400 ISP network 500 Internet 1000 Drive device 1001 Recording medium 1002 Auxiliary storage device 1003 Memory device 1004 CPU 1005 Interface device 1006 Display device 1007 Input device 1008 Output device

Claims

1. A graph generation method executed by a graph generation device, comprising: an acquisition step of acquiring traffic data of a host to be analyzed; and a graph generation step of dividing the traffic data into time slots, generating a graphlet for each time slot from the divided traffic data for each time slot, and generating a communication behavior graph in which communication behaviors corresponding to the graphlets are nodes and transitions between communication behaviors are edges.

2. The graph generating method according to claim 1, wherein in the graph generating step, the graph generating device determines a communication action corresponding to the graphlet based on a distance between a basic graphlet to which a communication action is assigned and the graphlet.

3. The graph generation method according to claim 2, wherein in the graph generation step, the graph generation device determines, as the communication action of the graphlet, the communication action of a basic graphlet that has the smallest distance between the graphlet and the basic graphlet in a set of basic graphlets.

4. The graph generation method according to claim 1, wherein in the graph generation step, the graph generation device generates edges for all communication behavior from a communication behavior in a certain time slot to a time slot a predetermined time after the certain time slot, and generates the communication behavior graph by deleting all edges between nodes when the number of overlapping edges between nodes is less than a predetermined number.

5. A graph generation device comprising: an acquisition unit that acquires traffic data of a host to be analyzed; and a graph generation unit that divides the traffic data into time slots, generates a graphlet for each time slot from the divided traffic data for each time slot, and generates a communication behavior graph in which communication behaviors corresponding to the graphlets are nodes and transitions between communication behaviors are edges.

6. A program for causing a computer to execute each step of the graph generating method according to any one of claims 1 to 4.

Citation Information

Patent Citations

  • Extensible network attack detection method based on behavior graph and Weisfeiler Lehman algorithm

    CN114900362A