Network anomaly detection method based on process flow comprehensive analysis
By combining the real-time detection of terminal stand-alone traffic and multi-terminal traffic within the network jurisdiction, combined with five-tuple and graph similarity detection, the lag and misjudgment problems of APT attack detection in the existing technology are solved, and the APT attack is realized Multi-dimensional observation and analysis improve detection accuracy and efficiency.
Patent Information
- Application Number
- CN202411112961.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-14
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-08-14
AI Technical Summary
In the prior art, APT attack detection cannot meet the actual network security defense needs through point-to-point and five-tuple-based network anomaly detection, resulting in delayed detection time, inefficient efficiency, and prone to misjudgment, and it is impossible to deeply mine the network flows. Related relationships, inability to accurately detect hidden cyber attacks.
Using a network abnormality detection method based on comprehensive process traffic analysis, the abnormality detection in the terminal is combined with abnormal behavior detection within the global network range, and the network traffic connection and interaction behavior is described from a macro perspective, through five-tuple and byte data of network traffic The spatial and temporal correlation detection of matching, pattern matching and graphs is used to perform similarity detection using Weisfeiler-Lehman graph kernel calculation method, and a dynamic time window is set for real-time analysis.
It improves detection accuracy and secure operation efficiency, can identify the complete link of APT attack, enhances the coverage of detection range, reduces the false alarm rate, and can capture slight changes in APT attacks at different time points.
Smart Images

Figure CN119011250A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of digital information transmission, in particular to a network anomaly detection method based on comprehensive analysis of process traffic. Background Art
[0002] The business scenarios in the information and communication network environment are numerous and complex. The iterative development of attack methods has made network attacks more targeted, covert and persistent. The attack steps are carefully planned and designed.
[0003] For example, advanced threat attacks APT will use terminal process injection to hide their malicious behaviors using normal business white processes. APT attacks are a multi-stage compound attack method with complex and diverse attack types, and many behaviors only show slight changes on a single network flow, so it is difficult to directly detect and judge them based on a certain feature. Moreover, even if it is the same APT attack, when launching attacks on different target networks, it will take more targeted strategies according to local conditions, making it difficult to defend against.
[0004] Therefore, the existing technology for APT attack detection, through point-to-point, firewall and five-tuple-based network anomaly detection, can no longer meet the actual network security defense needs. At present, attack behavior detection based on comprehensive analysis and judgment of security operations is performed by security analysts based on a large amount of alarm information and original traffic information. The troubleshooting time is as short as 10 minutes and as long as hours. The processing time is delayed, the efficiency is low, and it is easy to make misjudgments. In addition, only analyzing the characteristics of a single network flow will cause a large amount of detailed information loss, making it difficult to capture the correlation between network flows and unable to deeply mine network abnormal behaviors. Therefore, it is impossible to give accurate detection results when hidden network attacks come.
[0005] Based on the above reasons, the present invention designs a network anomaly detection method based on comprehensive analysis of process traffic, which combines anomaly detection within the terminal with abnormal behavior detection within the global network, comprehensively and accurately describes network traffic connections and interactive behaviors from a macro perspective, and improves detection accuracy. Summary of the invention
[0006] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a network anomaly detection method based on comprehensive analysis of process traffic, which combines anomaly detection within the terminal with abnormal behavior detection within the global network, comprehensively and accurately describes network traffic connections and interactive behaviors from a macro perspective, and improves detection accuracy.
[0007] In order to achieve the above object, the present invention provides a network anomaly detection method based on process traffic comprehensive analysis, comprising the following steps: S1, real-time detection of network anomalies based on terminal single-machine traffic, including: S1-1, detection based on network traffic quintuple: Based on a clear understanding of network attack behaviors, APT attack detection experts directly provide anomaly detection rules based on past experience or standards; S1-2, detection based on network traffic byte data matching: Using the information extracted when the function is called or returned collected by the data acquisition system module as input, regular expression matching is performed; S1-3, filtering based on pattern matching: Specify rule filtering scripts, support Python scripts, LUA scripts, FDS scripts, NASL scripts and PIT scripts; S2, real-time detection of network anomalies based on all terminal traffic within the network domain, including: S2-1, time series based detection: S2-1-1, set a network anomaly detection time window α, and sort out the network access behavior chain of each terminal within the network jurisdiction based on the collected terminal single-machine traffic within the α time period; S2-1-2: traverse and analyze all terminals within the network domain to see if they have the same network access behavior chain, and associate and classify the same behaviors; S2-2, detection based on spatial sequence: At a certain instant, a network behavior interaction graph is constructed based on the collected network five-tuple data. Each IP address is treated as an entity node, the association between the source IP address and the destination IP address is used as an edge, and the data packet sending direction of the source IP address and the destination IP address is used as the direction of the edge, thereby constructing a directed unweighted graph based on network behavior interaction. S2-3, Graph-based spatiotemporal correlation detection: S2-3-1, perform similarity detection on the directed unweighted graph based on network behavior interaction within the α time period: divide the α time period into t smaller time periods on average, then B=(β0,β1,β2,…,β t ), the directed unweighted graph at time β0 is sequentially connected with β1, β2, …, β t Calculate graph similarity of the directed unweighted graph at the time; S2-3-2, sequentially connect the directed unweighted graph at time β1 with β2, β3, …, β t The directed unweighted graph at the time is used to calculate the graph similarity until β is completed. t-1 Moment and Beta t Calculate the similarity of the directed unweighted graph at each moment and sort the calculation results; S2-3-3, sort the calculation results, and the APT attack detection expert sets the abnormal behavior threshold to detect abnormal situations.
[0008] The information extracted when a function is called or returned in S1-2 is one or several continuous segments of memory.
[0009] The network access behavior chain objects in S2-1-1 and S2-1-2 include the target host IP, target host port number and target host data.
[0010] The similarity calculation in S2-3-2 adopts a calculation method based on the Weisfeiler-Lehman kernel.
[0011] Compared with the prior art, the present invention combines the real-time detection of terminal single-machine traffic and multi-terminal traffic within the network jurisdiction, and at the same time improves the understanding and analysis capabilities of security events based on the host process chain, discovers the complete link of attack actions, and improves security operation efficiency. The present invention realizes multi-dimensional observation and analysis of APT attacks and improves the coverage capability of the detection range. Combining the abnormal detection in the terminal with the abnormal behavior detection within the global network, the perception ability is better and can accurately describe the network traffic connection and interactive behavior. The idea of the graph can cross-check and reduce false alarms. Using the calculation method based on the Weisfeiler-Lehman graph kernel for graph similarity detection, it is possible to identify abnormal patterns in the network behavior interaction graph, and even if the APT attack shows slight changes at different time points, it can be effectively captured and identified.
[0012] The present invention sets a dynamic time window α, and sorts out and classifies network behaviors within the window for correlation analysis. It can track and analyze the evolution of network behaviors in real time, combat APT behaviors in a timely manner in terms of persistence and complexity, and ensure detection timeliness, which is superior to existing solutions. BRIEF DESCRIPTION OF THE DRAWINGS
[0013] Figure 1 It is a schematic diagram of the detection method of the present invention. DETAILED DESCRIPTION
[0014] The present invention will now be further described with reference to the accompanying drawings.
[0015] join Figure 1 , the present invention provides a network anomaly detection method based on comprehensive analysis of process traffic: The following steps are involved: S1, real-time detection of network anomalies based on terminal single-machine traffic, including: S1-1, detection based on network traffic quintuple: Based on a clear understanding of network attack behaviors, APT attack detection experts directly provide anomaly detection rules based on past experience or standards; S1-2, detection based on network traffic byte data matching: Using the information extracted when the function is called or returned collected by the data acquisition system module as input, regular expression matching is performed; S1-3, filtering based on pattern matching: Specify rule filtering scripts, support Python scripts, LUA scripts, FDS scripts, NASL scripts and PIT scripts; S2, real-time detection of network anomalies based on all terminal traffic within the network domain, including: S2-1, time series based detection: S2-1-1, set a network anomaly detection time window α, and sort out the network access behavior chain of each terminal within the network jurisdiction based on the collected terminal single-machine traffic within the α time period; S2-1-2: traverse and analyze all terminals within the network domain to see if they have the same network access behavior chain, and associate and classify the same behaviors; S2-2, detection based on spatial sequence: At a certain instant, a network behavior interaction graph is constructed based on the collected network five-tuple data. Each IP address is treated as an entity node, the association between the source IP address and the destination IP address is used as an edge, and the data packet sending direction of the source IP address and the destination IP address is used as the direction of the edge, thereby constructing a directed unweighted graph based on network behavior interaction. S2-3, Graph-based spatiotemporal correlation detection: S2-3-1, perform similarity detection on the directed unweighted graph based on network behavior interaction within the α time period: divide the α time period into t smaller time periods on average, then B=(β0,β1,β2,…,β t ), the directed unweighted graph at time β0 is sequentially connected with β1, β2, …, β t Calculate graph similarity of the directed unweighted graph at the moment; S2-3-2, sequentially connect the directed unweighted graph at time β1 with β2, β3, …, β t The directed unweighted graph at the time is used to calculate the graph similarity until β is completed. t-1 Moment and Beta t Calculate the similarity of the directed unweighted graph at each moment and sort the calculation results; S2-3-3, sort the calculation results, and the APT attack detection expert sets the abnormal behavior threshold to detect abnormal situations.
[0016] The information extracted when a function is called or returned in S1-2 is one or several continuous segments of memory.
[0017] The network access behavior chain objects in S2-1-1 and S2-1-2 include the target host IP, target host port number and target host data.
[0018] The similarity calculation in S2-3-2 adopts a calculation method based on the Weisfeiler-Lehman kernel.
[0019] The above are only preferred embodiments of the present invention, which are only used to help understand the method and core ideas of the present application. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.
[0020] The present invention solves the problem that the existing technology for APT attacks through point-to-point, firewall and quintuple-based network anomaly detection can no longer meet the actual network security defense needs, resulting in unstable detection time, easy misjudgment, inability to conduct in-depth mining and thus inability to provide accurate detection results. The present invention combines the real-time detection of terminal single-machine traffic and multi-terminal traffic within the network jurisdiction, and at the same time improves the understanding and analysis capabilities of security events based on the host process chain, discovers the complete link of attack actions, and improves security operation efficiency. The present invention realizes multi-dimensional observation and analysis of APT attacks and improves the coverage capability of the detection range.
Claims
1. A network anomaly detection method based on comprehensive analysis of process traffic, characterized in that: The following steps are involved: S1, real-time detection of network anomalies based on terminal single-machine traffic, including: S1-1, detection based on network traffic quintuple: Based on a clear understanding of network attack behaviors, APT attack detection experts directly provide anomaly detection rules based on past experience or standards; S1-2, detection based on network traffic byte data matching: Using the information extracted when the function is called or returned collected by the data acquisition system module as input, regular expression matching is performed; S1-3, filtering based on pattern matching: Specify rule filtering scripts, support Python scripts, LUA scripts, FDS scripts, NASL scripts and PIT scripts; S2, real-time detection of network anomalies based on all terminal traffic within the network domain, including: S2-1, time series based detection: S2-1-1, set a network anomaly detection time window α, and sort out the network access behavior chain of each terminal within the network jurisdiction based on the collected terminal single-machine traffic within the α time period; S2-1-2: traverse and analyze all terminals within the network domain to see if they have the same network access behavior chain, and associate and classify the same behaviors; S2-2, detection based on spatial sequence: At a certain instant, a network behavior interaction graph is constructed based on the collected network five-tuple data. Each IP address is treated as an entity node, the association between the source IP address and the destination IP address is used as an edge, and the data packet sending direction of the source IP address and the destination IP address is used as the direction of the edge, thereby constructing a directed unweighted graph based on network behavior interaction. S2-3, Graph-based spatiotemporal correlation detection: S2-3-1, perform similarity detection on the directed unweighted graph based on network behavior interaction within the α time period: divide the α time period into t smaller time periods on average, then B=(β0,β1,β2,…,β t ), the directed unweighted graph at time β0 is sequentially connected with β1, β2, …, β t Calculate graph similarity of the directed unweighted graph at the moment; S2-3-2, sequentially connect the directed unweighted graph at time β1 with β2, β3, …, β t The directed unweighted graph at the time is used to calculate the graph similarity until β is completed. t-1 Moment and Beta t Calculate the similarity of the directed unweighted graph at each moment and sort the calculation results; S2-3-3, sort the calculation results, and the APT attack detection expert sets the abnormal behavior threshold to detect abnormal situations.
2. The network anomaly detection method based on comprehensive analysis of process traffic according to claim 1 is characterized in that: The information extracted when the function in S1-2 is called or returned is one or several continuous memory segments.
3. The network anomaly detection method based on process traffic comprehensive analysis according to claim 1 is characterized in that: The network access behavior chain objects in S2-1-1 and S2-1-2 include the target host IP, the target host port number and the target host data.
4. The network anomaly detection method based on process traffic comprehensive analysis according to claim 1 is characterized in that: The similarity calculation in S2-3-2 adopts a calculation method based on the Weisfeiler-Lehman kernel.
Citation Information
Patent Citations
Host network behavior mode measurement method based on attribute relation graph
CN109040130A
Traceability graph APT attack real-time intrusion detection method of deep echo network
CN117544377A
User and entity behavioral analysis with network topology enhancements
US20220060497A1
Cited By
Intranet traffic-oriented security detection method, apparatus and device, and medium
CN122137604A
