Multi-dimensional feature-based malicious sample analysis methods, devices, and systems

Through the multi-dimensional feature collaborative analysis method and contribution weight aggregation technology, combined with the secret sharing mechanism, the problems of low detection accuracy and poor scalability in the existing malicious sample detection methods are solved, and the accuracy and flexibility of malicious sample detection are achieved. .

CN118862064BActive Publication Date: 2024-12-03HANGZHOU HIKVISION DIGITAL TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411320858.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-20
Publication Date
2024-12-03
Estimated Expiration
2044-09-20

AI Technical Summary

Technical Problem

The existing malicious sample detection methods are based on a single feature, resulting in low detection accuracy and prone to errors or failures, and cannot effectively deal with the problems of diverse malicious sample shapes and multi-dimensional discrete features.

Method used

Using multi-dimensional feature collaboration analysis method, the detection engine corresponding to different detection methods nodes in the detection domain hierarchical structure detects the feature data of different dimensions in the current sample, and uses the contribution weight to gather results, dynamically adjust the weight of the detection results to improve the Detection accuracy, combined with a secret sharing mechanism to control forgery and attacks of the detection engine.

Benefits of technology

It improves the scalability and detection accuracy of malicious sample detection, solves the problems of unscalable detection dimensions and low detection accuracy, and adapts to different detection engines and other problems such as inconsistent capabilities and failures, and realizes flexible detection methods.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118862064B_ABST
    Figure CN118862064B_ABST
Patent Text Reader

Abstract

The embodiments of the present application provide a method, device and system for analyzing malicious samples with multi-dimensional feature collaboration. With the help of the detection domain hierarchical structure, the present embodiment effectively uses the detection engines corresponding to different detection method nodes in the detection domain hierarchical structure to detect the feature data of different dimensions in the current sample to analyze whether the current sample is a malicious sample, which realizes the collaboration of feature data in multiple dimensions for malicious sample analysis.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of information security, and in particular to a method, device and system for analyzing malicious samples with multi-dimensional feature collaboration. Background Art

[0002] Existing malicious sample detection methods are usually based on single features such as file features (such as file hash, file code structure, etc.) or behavioral features (kernel call sequence after code execution, file system change features, traffic behavior features, etc.). Specifically, when performing malicious sample detection, the above features can be extracted from a labeled data set, and then an algorithm model such as machine learning or deep learning can be trained based on the extracted features, and then malicious sample detection can be performed based on the algorithm model.

[0003] However, the above single feature often limits the accuracy of malicious sample detection, and may even cause detection errors or failures. Summary of the invention

[0004] The embodiments of the present application provide a method, device and system for malicious sample analysis with multi-dimensional feature collaboration to implement malicious sample analysis with multi-dimensional feature collaboration.

[0005] The embodiment of the present application provides a method for analyzing malicious samples with multi-dimensional feature collaboration, the method being applied to a central server, and the method comprising:

[0006] Determine the target sample type node to which the current sample belongs from the constructed detection domain hierarchical structure; the detection domain hierarchical structure includes at least one sample type node, at least one dimension node connected under each sample type node, and at least one detection method node connected under each dimension node; any node in the detection domain hierarchical structure is assigned a contribution weight for malicious sample analysis;

[0007] For each dimension node connected to the target sample type node, based on the dimension feature required by the detection method indicated by each detection method node under the dimension node, the dimension feature data corresponding to the dimension feature is extracted from the current sample and sent to the detection engine running the detection method for detection processing to obtain the detection result corresponding to the current sample; the detection engine is assigned a participation weight for participating in malicious sample analysis of samples belonging to the target sample type; the target sample type refers to the sample type indicated by the target sample type node;

[0008] Receive the detection result corresponding to the current sample reported by the detection engine; if the participation weight assigned to each detection engine that has reported the detection result meets the requirements of the set secret sharing method, then for each detection result reported by the detection engine, determine the contribution weight of the detection result based on the contribution weight assigned to the detection method node corresponding to the detection method run by the detection engine, the contribution weight assigned to the dimension node connected to the detection method node, and the contribution weight assigned to the target sample type node; aggregate each detection result according to the contribution weight of the detection result to analyze whether the current sample is a malicious sample based on the aggregated result.

[0009] This embodiment provides a multi-dimensional feature collaborative malicious sample analysis system, the system comprising a central server and at least one edge device independent of the central server;

[0010] The central server executes the steps in the method described above;

[0011] Any edge device loads at least one detection engine registered with the central server; any detection engine runs a detection method indicated by a detection method node, and the detection engine is assigned a participation weight for participating in malicious sample analysis of samples belonging to a target sample type; the target sample type refers to the sample type indicated by the target sample type node; the target sample type node is the sample type node to which the current sample belongs, determined by the central server from the constructed detection domain hierarchical structure according to the steps in the method as described above.

[0012] This embodiment provides a malicious sample analysis device with multi-dimensional feature collaboration, which is applied to a central server and includes:

[0013] A determination unit, used to determine the target sample type node to which the current sample belongs from a constructed detection domain hierarchical structure; the detection domain hierarchical structure includes at least one sample type node, at least one dimension node connected under each sample type node, and at least one detection method node connected under each dimension node; any node in the detection domain hierarchical structure is assigned a contribution weight for malicious sample analysis; and,

[0014] For each dimension node connected to the target sample type node, based on the dimension feature required by the detection method indicated by each detection method node under the dimension node, the dimension feature data corresponding to the dimension feature is extracted from the current sample and sent to the detection engine running the detection method for detection processing to obtain the detection result corresponding to the current sample; the detection engine is assigned a participation weight for participating in malicious sample analysis of samples belonging to the target sample type; the target sample type refers to the sample type indicated by the target sample type node;

[0015] An analysis unit is used to receive a detection result corresponding to a current sample reported by a detection engine; if the participation weight assigned to each detection engine that has reported the detection result meets the requirements of a set secret sharing method, then for each detection result reported by the detection engine, based on the contribution weight assigned to the detection method node corresponding to the detection method run by the detection engine, the contribution weight assigned to the dimension node connected to the detection method node, and the contribution weight assigned to the target sample type node, determine the contribution weight of the detection result; aggregate each detection result according to the contribution weight of the detection result to analyze whether the current sample is a malicious sample based on the aggregated result.

[0016] It can be seen from the above technical solutions that this embodiment effectively utilizes the detection engines corresponding to different detection method nodes in the detection domain hierarchical structure with the help of the detection domain hierarchical structure to detect the feature data in different dimensions in the current sample to analyze whether the current sample is a malicious sample. This realizes the collaborative analysis of malicious samples using feature data in multiple dimensions.

[0017] Furthermore, in this embodiment, when analyzing whether the current sample is a malicious sample, the contribution weight of the detection result reported by each detection engine will be determined based on the contribution weight assigned to the detection method node corresponding to the detection method run by the detection engine, the contribution weight assigned to the dimension node connected to the detection method node, and the contribution weight assigned to the target sample type node; then, based on the contribution weight of each detection result, each detection result is aggregated to analyze whether the current sample is a malicious sample based on the aggregated result, instead of directly using the fixed weight of a certain detection engine or a certain detection method as the contribution weight of the detection result reported by the detection engine. In essence, the contribution weight of the detection result reported by the detection engine is dynamically updated with the detection domain hierarchical structure and the specific detection sample (such as the target sample type node to which it belongs). This adaptive adjustment of the contribution weights of the detection results reported by the above-mentioned detection engines based on the detection sample can solve the problems of low detection accuracy caused by the diverse forms of malicious samples, multi-dimensional discrete features, and non-scalability of detection dimensions, and effectively improves the scalability and detection accuracy of malicious sample detection.

[0018] Furthermore, this embodiment, with the help of a secret sharing mechanism, can effectively control the forgery of each detection engine and internal attacks on the detection engine, and can adapt to problems such as inconsistent detection time, inconsistent detection capabilities, engine failure or no message offline of different detection engines. There is no need to continuously wait for engine detection results, and the detection method is more flexible. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the specification and, together with the description, serve to explain the principles of the specification.

[0020] Figure 1 A flow chart of the method provided in the embodiment of the present application;

[0021] Figure 2 A diagram of the hierarchical structure of the detection domain provided in the embodiment of the present application;

[0022] Figure 3 A registration flow chart provided for an embodiment of the present application;

[0023] Figure 4 A system structure diagram provided for an embodiment of the present application;

[0024] Figure 5 A structural diagram of a device provided in an embodiment of the present application;

[0025] Figure 6 A structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the device structure diagram provided in the embodiments of the present application; understand the technical scheme provided in the embodiments of the present application, and make the above-mentioned purposes, features and advantages of the embodiments of the present application more obvious and easy to understand, the technical scheme in the embodiments of the present application is further described in detail below with reference to the accompanying drawings.

[0027] See also Figure 1 , Figure 1 A flow chart of a method provided in an embodiment of the present application. The method is applied to a central server.

[0028] like Figure 1 As shown, the process may include the following steps:

[0029] Step 101 : Determine the target sample type node to which the current sample belongs from the constructed detection domain hierarchy structure.

[0030] The current sample here refers to the sample that needs to be analyzed for malicious samples.

[0031] In specific applications, malicious samples of different sample types or malicious samples of the same sample type have different dimensional features in different dimensions (referred to as dimensional feature data).

[0032] Optionally, the sample types mentioned above are, for example, executable files, packet capture (PCAP: Packet Capture) files (which are file types ending with .pcap and contain data flow information), etc., which are not specifically limited in this embodiment.

[0033] Optionally, any of the above sample types can be described by at least one dimension. The dimensions under any sample type can be set according to actual conditions, as long as they can describe the characteristics of the sample type. For example, if the sample type is a PCMP data packet, the dimensions under the sample type may include: data packet statistical features (such as variance values ​​obtained by performing variance calculations on the sizes of each data packet counted over a period of time, etc.), data packet character features (such as variance values ​​obtained by performing variance calculations on the number of characters in each data packet counted over a period of time, etc.), pixel-level features after data packet conversion to an image (such as the number of data packets under each session flow over a period of time converted into a vector according to binary, etc.), etc. For another example, if the sample type is an executable file, the dimensions under the sample type may include: the hash value of the executable file, the executable file code structure, the system call sequence of the dynamic execution of the executable file, etc.

[0034] In this embodiment, any dimension has at least one detection method (denoted as E) for detecting feature data in the dimension. For example, if the dimension is the above-mentioned statistical feature of the data packet, the detection method in the dimension may be a detection method such as machine learning and logistic regression.

[0035] Based on the above description, this embodiment can construct a detection domain hierarchical structure based on sample type, dimensions under sample type, detection methods under each dimension, etc. As an embodiment, the detection domain hierarchical structure here includes at least: at least one sample type node (indicating its corresponding sample type), and at least one dimension node connected to each sample type node (indicating its corresponding dimension), and at least one detection method node connected to each dimension node (indicating at least one detection method under the dimension corresponding to the dimension node). Figure 2 The above detection domain hierarchical structure is shown as an example. Figure 2 As shown in FIG. , the detection domain hierarchy structure here can be a tree structure. Figure 2 In the above formula, DA = {DB1, DB2, DB3…DBn}. DA refers to the sample detection domain divided by sample type, and DB refers to the dimension node when the sample type is determined. Any DB can be represented by the following formula: {E1, E2, E3…Em}. E refers to the specific detection method node when the dimension is determined.

[0036] Based on the detection domain hierarchical structure described above, in this step 101, determining the target sample type node to which the current sample belongs from the constructed detection domain hierarchical structure may include: determining the sample type of the current sample; and determining the target sample type node corresponding to the sample type from the constructed detection domain hierarchical structure.

[0037] Step 102, for each dimensional node connected to the target sample type node, based on the dimensional features required by the detection method indicated by each detection method node under the dimensional node, extract the dimensional feature data corresponding to the dimensional feature from the current sample and send it to the detection engine running the detection method for detection processing to obtain the detection result corresponding to the current sample.

[0038] As an embodiment, the detection engine here may run on the above-mentioned central server, and may also run on other devices different from the above-mentioned central server, such as edge devices managed by the central server.

[0039] In this embodiment, different detection methods may require different feature data. For each detection method node, this embodiment adaptively extracts the dimensional feature data corresponding to the dimensional feature from the current sample based on the dimensional feature required by the detection method indicated by the detection method node, such as the statistical variance of the data packet, so as to improve the detection accuracy of the detection engine.

[0040] As an embodiment, before sending the dimensional feature data to the detection engine in step 102, the detection engine needs to be registered with the above-mentioned central server first. Figure 3 The process of registering the detection engine with the above-mentioned central server is illustrated by way of example. During the process of registering with the above-mentioned central server, the detection engine will inform the central server of the sample types, dimensional features, etc. that it supports, so that when executing step 102, the dimensional feature data corresponding to the dimensional feature is adaptively extracted from the current sample and sent to the detection engine, so as to improve the detection accuracy of the detection engine.

[0041] Step 103, receiving the detection result corresponding to the current sample reported by the detection engine; if the participation weight assigned to each detection engine that has reported the detection result meets the requirements of the set secret sharing method, then for each detection result reported by the detection engine, based on the contribution weight assigned to the detection method node corresponding to the detection method run by the detection engine, the contribution weight assigned to the dimension node connected to the detection method node, and the contribution weight assigned to the target sample type node, determine the contribution weight of the detection result; and aggregate each detection result according to the contribution weight of the detection result to analyze whether the current sample is a malicious sample based on the aggregated result.

[0042] In this embodiment, a secret sharing method is introduced. Based on the secret sharing method, each detection engine is assigned a participation weight for participating in malicious sample analysis of samples belonging to a target sample type. The target sample type here refers to the sample type indicated by the target sample type node.

[0043] As an embodiment, the sum of the participation weights of the detection engines running on the central server that participate in the malicious sample analysis of the samples belonging to the above target sample type is less than the set minimum participation weight. Taking the minimum participant weight of 0.9 as an example, for example, the sum of the participation weights of the detection engines running on the central server that participate in the malicious sample analysis of the samples belonging to the above target sample type is 0.8. The sum of the participation weights of the other detection engines that are not on the central server and participate in the malicious sample analysis of the samples belonging to the above target sample type is 0.2.

[0044] In specific implementation, the detection results reported by each detection engine may have a certain sequence, and there may even be a detection engine failure that results in no detection result being returned. In this case, in order to avoid the situation where different detection engines report detection results for different lengths of time (also called detection time) and different detection capabilities of different detection engines cause the central server to continue waiting for the detection engine to return the detection results, this embodiment, with the help of the above-mentioned secret sharing method, does not require the central server to continue waiting for the detection engine to return the detection results. Instead, when it is found that the participation weights assigned to each detection engine that has reported the detection results meet the set secret sharing method requirements, for example, the sum of the participation weights assigned to each detection engine that has reported the detection results is greater than or equal to the set minimum participation weight, then the operation of aggregating the detection results reported by each detection engine in step 103 can be started.

[0045] As an embodiment, the above detection result may be, for example, a first value such as 0 or a second value such as 1, wherein the first value indicates malicious intent and the second value indicates non-malicious intent.

[0046] It should be noted that, in this embodiment, any node in the detection domain hierarchy is assigned a contribution weight for malicious sample analysis. For example, the detection method node is assigned a contribution weight for malicious sample analysis, the dimension node connected to the detection method node is assigned a contribution weight for malicious sample analysis, and the sample type node connected to the dimension node is assigned a contribution weight for malicious sample analysis.

[0047] Optionally, in a specific implementation, malicious sample analysis of the current sample is triggered based on an event. The event here is, for example, an external instruction, or it can be the detection of a suspected malicious sample. Malicious sample analysis can be performed on at least one sample under the same event. Correspondingly, the above-mentioned detection domain hierarchy structure also includes: an event node at the upper layer of the sample type node. The same event node can be connected to at least one sample type node. Under this premise, from bottom to top, the contribution weight assigned to the detection method node for malicious sample analysis, the contribution weight assigned to the dimension node connected to the detection method node for malicious sample analysis, and the contribution weight assigned to the sample type node connected to the dimension node for malicious sample analysis are all contributions to the malicious sample analysis of the event represented by the event node connected to the sample type node. Under this premise, optionally, the contribution weight assigned to the event node for malicious sample analysis can be 1.

[0048] As an embodiment, the sum of the contribution weights of all sample type nodes under the same event node, the contribution weights of each dimension node connected to each sample type node, and the contribution weights of each detection method node connected to each dimension node is a specified value, such as 1.

[0049] Based on this, in the process of executing the aggregation operation in step 103, as described in step 103, first for each detection result reported by the detection engine, based on the contribution weight (recorded as WE) assigned to the detection method node corresponding to the detection method run by the detection engine, the contribution weight (recorded as WDB) assigned to the dimension node connected to the detection method node (i.e., the detection method node corresponding to the detection method run by the detection engine), and the contribution weight (recorded as WDA) assigned to the target sample type node, the contribution weight of the detection result is determined. That is, this embodiment does not directly use the fixed weight of a certain detection engine or a certain detection method as the contribution weight of the detection result reported by the detection engine, and it also dynamically updates the contribution weight of the detection result reported by the detection engine with the detection domain hierarchical structure and the specific detection sample (such as the target sample type node to which it belongs). This adaptive adjustment of the contribution weight of the detection results reported by the above detection engines based on the detection sample can solve the problems of low detection accuracy caused by the diverse forms of malicious samples and multi-dimensional discrete features, and the inscalability of detection dimensions, and effectively improves the scalability and detection accuracy of malicious sample detection.

[0050] As an embodiment, the contribution weight assigned to the detection method node corresponding to the detection method run by the detection engine, the contribution weight assigned to the dimension node connected to the detection method node, and the contribution weight assigned to the target sample type node can be subjected to a specified operation such as a multiplication operation to obtain an operation result, and the operation result is determined as the contribution weight of the detection result. Taking the specified operation as a multiplication operation as an example, the contribution weight of the detection result here can be WE*WDB*WDA.

[0051] Afterwards, the detection results are aggregated according to the contribution weights of the detection results to analyze whether the current sample is a malicious sample based on the aggregated results. Optionally, the detection results are aggregated according to the contribution weights of the detection results, for example, the contribution weights of the detection results can be multiplied by the product of the detection results to perform an addition operation, which is the above-mentioned aggregation.

[0052] As an embodiment, there are many ways to implement whether the current sample is a malicious sample based on the aggregation result analysis in specific implementation, for example, checking whether the aggregation result is greater than or equal to a set threshold. If so, it is determined that the current sample is not a malicious sample, otherwise, it is determined that the current sample is a malicious sample.

[0053] So far, completed Figure 1 The process shown.

[0054] pass Figure 1 It can be seen from the shown process that this embodiment, with the help of the detection domain hierarchical structure, effectively uses the detection engines corresponding to different detection method nodes in the detection domain hierarchical structure to detect feature data of different dimensions in the current sample to analyze whether the current sample is a malicious sample. This realizes the collaborative analysis of malicious samples with feature data in multiple dimensions.

[0055] Furthermore, in this embodiment, when analyzing whether the current sample is a malicious sample, the contribution weight of the detection result will be determined based on the detection result reported by each detection engine, the contribution weight assigned to the detection method node corresponding to the detection method run by the detection engine, the contribution weight assigned to the dimension node connected to the detection method node, and the contribution weight assigned to the target sample type node; then, the detection results are aggregated based on the contribution weights of each detection result to analyze whether the current sample is a malicious sample based on the aggregated results, rather than directly using the detection results reported by each detection engine for analysis, so that the detection accuracy is high and the detection capability is stronger.

[0056] Furthermore, this embodiment, with the help of a secret sharing mechanism, can effectively control the forgery of each detection engine and internal attacks on the detection engine, and can adapt to problems such as inconsistent detection time, inconsistent detection capabilities, engine failure or no message offline of different detection engines. There is no need to continuously wait for engine detection results, and the detection method is more flexible.

[0057] The following describes the registration process of the detection engine:

[0058] In this embodiment, the central server opens a communication interface, and each detection engine (which may belong to different manufacturers, different owners, and different regions) actively registers with the central server.

[0059] like Figure 3 As shown, the registration process may include the following steps:

[0060] Step 301: The detection engine sends a registration message to the central server.

[0061] Step 302: The central server sends a registration request message.

[0062] As an embodiment, the registration requirement information here is, for example, basic information of the detection engine such as IP address, engine name, detection domain information (such as sample types supported for detection, dimensional features, etc.).

[0063] Step 303: The detection engine reports the information required by the central server.

[0064] Step 304: The central server assigns a participation weight Wkey to the detection engine. At this point, the detection engine completes registration.

[0065] As described above, the participation weight Wkey here does not participate in the calculation of the final result, but is only used to determine whether the overall secret sharing weight reaches the minimum participant weight.

[0066] So far, completed Figure 3 The registration process of the detection engine is shown.

[0067] It should be noted that the participation weight of the above detection engine is not fixed, and it will be adjusted based on subsequent loyalty tests of the detection engine.

[0068] As an embodiment, for each detection engine, the loyalty of the detection engine may be detected based on at least one of the following methods:

[0069] 1) Detect the loyalty of the detection engine based on the engine behavior.

[0070] For example, the loyalty of the detection engine is determined based on the communication behavior between the detection engine and the central server.

[0071] As an embodiment, the above-mentioned determining the loyalty of the detection engine based on the communication behavior between the detection engine and the central server may include: when the detection engine is not in the central server, performing the following judgment:

[0072] Determine whether the detection result reported by the detection engine is not received within a set time period after the dimensional feature data is sent to the detection engine; and / or, on the premise that the detection result reported by the detection engine is received within the set time period, determine whether the session established between the detection engine and the central server based on the participation weight of the detection engine is erroneous; determine the loyalty of the detection engine based on the judgment result, wherein the loyalty of the detection engine is lower when the judgment result is yes than when the judgment result is no.

[0073] In this embodiment, any detection engine will establish an encrypted channel with the central server based on its assigned participation weight Wkey, using the Wkey as the channel key, to ensure that data between the detection engine and the central server, such as the detection result reported by the detection engine, is not leaked. When the central server receives the detection result reported by the detection engine through the above-mentioned encrypted channel, if the decryption of the detection result based on the participation weight of the detection engine fails, it is determined that the session between the detection engine and the central server established based on the participation weight of the detection engine is wrong (the detection engine may be counterfeited). If the decryption is successful, it is determined that the above-mentioned session is successfully established.

[0074] When it is determined that the detection result reported by the detection engine is not received within the above-mentioned set time period, the detection time of the detection engine is too long (the detection engine may be faulty, may be offline, etc.) or too short (the detection engine may be counterfeited), then it can be considered that the loyalty of the detection engine is lower than the previous loyalty, that is, the loyalty of the detection engine is reduced; and / or,

[0075] When it is determined that the session between the detection engine and the central server established based on the participation weight of the detection engine is wrong, it is considered that the detection engine may be counterfeited. At this time, it can be considered that the loyalty of the detection engine is lower than the previous loyalty, that is, the loyalty of the detection engine is reduced.

[0076] 2) Detect the loyalty of the detection engine based on the detection results reported by the detection engine.

[0077] For example, the loyalty of the detection engine is determined based on the detection results reported by the detection engine.

[0078] In specific implementation, as an embodiment, the loyalty of the detection engine can be determined based on the comparison result between the detection result reported by the detection engine and the above-mentioned aggregation result; wherein, the greater the difference between the detection result reported by any detection engine and the aggregation result, the lower the loyalty of the detection engine.

[0079] As another embodiment, the loyalty of the detection engine can also be determined in accordance with a hierarchical comparison method. For example, for the dimension node connected to the detection method node corresponding to the detection method run by the detection engine, based on the contribution weights assigned to the detection method nodes connected to the dimension node, the detection results reported by the detection engines that respectively run the detection methods corresponding to the detection method nodes are aggregated to obtain the aggregated sub-result corresponding to the dimension node; the loyalty of the detection engines that run the detection methods corresponding to the detection method nodes connected to the dimension node is determined based on the aggregated sub-result; wherein, the greater the difference between the detection result reported by any detection engine and the aggregated sub-result, the lower the loyalty of the detection engine. This method of determining the loyalty of the detection engine in accordance with a hierarchical comparison method can be adjusted in a timely manner as other nodes in the detection domain hierarchical structure are adjusted, so as to dynamically update the participation weight of the detection engine as the detection process progresses.

[0080] 3) Detect the fidelity of the detection engine based on active testing.

[0081] For example, the loyalty of the detection engine is determined based on the test results of the detection engine. In specific implementation, the detection engine can be tested using pre-calibrated test samples, and the loyalty of the detection engine can be determined based on the test results; wherein, the greater the difference between the test result of any test sample and the label of the calibrated test sample, the lower the loyalty of the detection engine.

[0082] After the loyalty of the detection engine is determined by at least one of the above methods, the participation weight of the detection engine can be adjusted based on the loyalty of the detection engine; wherein, the smaller the loyalty of the detection engine, the smaller the participation weight of the detection engine. The smaller the participation weight of the detection engine, the less likely it is that the detection results reported by the detection engine will participate in the above-mentioned convergence, and vice versa. The more likely it is to participate in the above-mentioned convergence. This realizes the dynamic adjustment of the participation weight of any detection engine as the detection process progresses, so as to avoid the problem of detection errors or failures caused by malicious attacks by the detection engine.

[0083] Optionally, as an embodiment, when it is detected that the participation weight of any detection engine is less than a set weight threshold, the detection engine is deregistered, and the detection engine is prohibited from registering with the central server again within a set time period after deregistration.

[0084] Optionally, as an embodiment, the present embodiment may use the analysis result of whether the current sample is a malicious sample as the sample label of the current sample and add it together with the current sample as labeled data to the original data set; then optimize and train the detection method of each detection engine based on the original data set; and,

[0085] Adjust the contribution weight assigned to at least one node in the detection domain hierarchy, such as the detection method node corresponding to the optimized detection method. Of course, this embodiment also dynamically adjusts the contribution weights assigned to other nodes in the detection domain hierarchy except the detection method node according to current actual needs.

[0086] It should be noted that, in this embodiment, the contribution weight assigned to each node in the above-mentioned detection domain hierarchical structure can be set according to actual needs. For example, for the detection method node, it can be set based on the accuracy of the corresponding detection method. The higher the accuracy, the greater the contribution weight of the corresponding detection node.

[0087] The method provided in the embodiment of the present application is described above. The system and device provided in the embodiment of the present application are described below:

[0088] See also Figure 4 , Figure 4 This is a system structure diagram provided in the embodiment of the present application. Figure 4 As shown, the system includes a central server, and at least one edge device independent of the central server;

[0089] The central server executes the steps in the above method;

[0090] Any edge device loads at least one detection engine registered with the central server; any detection engine runs a detection method indicated by a detection method node, and the detection engine is assigned a participation weight for participating in malicious sample analysis of samples belonging to a target sample type; the target sample type refers to the sample type indicated by the target sample type node; the target sample type node is the sample type node to which the current sample belongs, determined by the central server from the constructed detection domain hierarchical structure in accordance with the steps in the above method.

[0091] As an embodiment, the present application also provides a malicious sample analysis device with multi-dimensional feature collaboration. Figure 5 , Figure 5 This is a diagram of the device structure provided in the embodiment of the present application. Figure 5 As shown, the device comprises:

[0092] A determination unit, used to determine the target sample type node to which the current sample belongs from a constructed detection domain hierarchical structure; the detection domain hierarchical structure includes at least one sample type node, at least one dimension node connected under each sample type node, and at least one detection method node connected under each dimension node; any node in the detection domain hierarchical structure is assigned a contribution weight for malicious sample analysis; and,

[0093] For each dimension node connected to the target sample type node, based on the dimension feature required by the detection method indicated by each detection method node under the dimension node, the dimension feature data corresponding to the dimension feature is extracted from the current sample and sent to the detection engine running the detection method for detection processing to obtain the detection result corresponding to the current sample; the detection engine is assigned a participation weight for participating in malicious sample analysis of samples belonging to the target sample type; the target sample type refers to the sample type indicated by the target sample type node;

[0094] An analysis unit is used to receive a detection result corresponding to a current sample reported by a detection engine; if the participation weight assigned to each detection engine that has reported the detection result meets the requirements of a set secret sharing method, then for each detection result reported by the detection engine, based on the contribution weight assigned to the detection method node corresponding to the detection method run by the detection engine, the contribution weight assigned to the dimension node connected to the detection method node, and the contribution weight assigned to the target sample type node, determine the contribution weight of the detection result; aggregate each detection result according to the contribution weight of the detection result to analyze whether the current sample is a malicious sample based on the aggregated result.

[0095] Optionally, for each detection result reported by the detection engine, based on the contribution weight assigned to the detection method node corresponding to the detection method run by the detection engine, the contribution weight assigned to the dimension node connected to the detection method node, and the contribution weight assigned to the target sample type node, determining the contribution weight of the detection result includes:

[0096] For each detection result reported by the detection engine, the contribution weight assigned to the detection method node corresponding to the detection method run by the detection engine, the contribution weight assigned to the dimension node connected to the detection method node, and the contribution weight assigned to the target sample type node are performed by performing specified operations to obtain the operation result, and the operation result is determined as the contribution weight of the detection result.

[0097] Optionally, the sum of the participation weights of the detection engines running on the central server and participating in malicious sample analysis of samples belonging to the target sample type is less than a set minimum participation weight;

[0098] The participation weights assigned to the detection engines that have reported the detection results meet the set secret sharing method requirements, which means that the sum of the participation weights assigned to the detection engines that have reported the detection results is greater than or equal to the set minimum participation weight.

[0099] Optionally, the analysis unit further comprises:

[0100] For each detection engine, determine the loyalty of the detection engine based on the communication behavior between the detection engine and the central server, and / or determine the loyalty of the detection engine based on the detection results reported by the detection engine, and / or determine the loyalty of the detection engine based on the test results of the detection engine;

[0101] Based on the loyalty of the detection engine, the participation weight of the detection engine is adjusted; wherein, the smaller the loyalty of the detection engine is, the smaller the participation weight of the detection engine is.

[0102] Optionally, determining the loyalty of the detection engine based on the communication behavior between the detection engine and the central server includes:

[0103] When the detection engine is not located on the central server, the following judgment is performed:

[0104] Determine whether the detection result reported by the detection engine is not received within a set time period after the dimensional feature data is sent to the detection engine; and / or, on the premise that the detection result reported by the detection engine is received within the set time period, determine whether the session between the detection engine and the central server established based on the participation weight of the detection engine is erroneous; wherein, when the central server receives the detection result reported by the detection engine through the encrypted channel established between the detection engine and the detection engine, if the decryption of the detection result based on the participation weight of the detection engine fails, it is determined that the session between the detection engine and the central server established based on the participation weight of the detection engine is erroneous; if the decryption is successful, it is determined that the session is successfully established; the encrypted channel is a secure channel constructed based on the participation weight of the detection engine as the channel key;

[0105] The loyalty of the detection engine is determined based on the judgment result, wherein when the judgment result is yes, the loyalty of the detection engine is lower than when the judgment result is no.

[0106] Optionally, determining the loyalty of the detection engine based on the detection result reported by the detection engine includes:

[0107] For the dimension node connected to the detection method node corresponding to the detection method run by the detection engine, based on the contribution weights assigned to the detection method nodes connected to the dimension node, the detection results reported by the detection engines that respectively run the detection methods corresponding to the detection method nodes are aggregated to obtain the aggregated sub-result corresponding to the dimension node; the loyalty of the detection engines that run the detection methods corresponding to the detection method nodes connected to the dimension node is determined based on the aggregated sub-result; wherein, the greater the difference between the detection result reported by any detection engine and the aggregated sub-result, the lower the loyalty of the detection engine; or,

[0108] For each detection engine, the loyalty of the detection engine is determined based on the comparison result between the detection result reported by the detection engine and the converged result; wherein, the greater the difference between the detection result reported by any detection engine and the converged result, the lower the loyalty of the detection engine.

[0109] Optionally, determining the loyalty of the detection engine based on a test result of the detection engine includes:

[0110] The detection engine is tested using pre-calibrated test samples, and the loyalty of the detection engine is determined based on the test results; wherein, the greater the difference between the test result of any test sample and the calibrated label of the test sample, the lower the loyalty of the detection engine.

[0111] Optionally, the analysis unit further comprises:

[0112] For each detection engine, if the participation weight of the detection engine is less than the set weight threshold, the detection engine is deregistered, and the detection engine is prohibited from registering with the central server again within a set time period after deregistration.

[0113] Optionally, the analysis unit further comprises:

[0114] The analysis result of whether the current sample is a malicious sample is used as the sample label of the current sample and added to the original data set together with the current sample as labeled data; the detection method for optimizing and training each detection engine based on the original data set; and

[0115] The contribution weight assigned to at least one node in the detection domain hierarchy is adjusted.

[0116] So far, completed Figure 5 Structural diagram of the device shown.

[0117] The present application also provides Figure 5 The hardware structure of the device shown. Figure 6 , Figure 6 This is a structural diagram of an electronic device provided in an embodiment of the present application. Figure 6 As shown, the hardware structure may include: a processor and a machine-readable storage medium, the machine-readable storage medium storing machine-executable instructions that can be executed by the processor; the processor is used to execute the machine-executable instructions to implement the method disclosed in the above example of this application.

[0118] Based on the same application concept as the above method, an embodiment of the present application also provides a machine-readable storage medium, on which a number of computer instructions are stored. When the computer instructions are executed by a processor, the method disclosed in the above example of the present application can be implemented.

[0119] Exemplarily, the above-mentioned machine-readable storage medium can be any electronic, magnetic, optical or other physical storage device, which can contain or store information, such as executable instructions, data, etc. For example, the machine-readable storage medium can be: RAM (Radom Access Memory), volatile memory, non-volatile memory, flash memory, storage drive (such as hard disk drive), solid state drive, any type of storage disk (such as optical disk, DVD, etc.), or similar storage medium, or a combination thereof.

[0120] The systems, devices, modules or units described in the above embodiments may be implemented by a computer processor or entity, or by a product having a certain function. A typical implementation device is a computer, and the specific form of the computer may be a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email transceiver device, a game console, a tablet computer, a wearable device or a combination of any of these devices.

[0121] For the convenience of description, the above device is described in various units according to their functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0122] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the embodiments of the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0123] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0124] Moreover, these computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0125] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0126] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A malicious sample analysis method based on multi-dimensional feature collaboration, characterized in that: The method is applied to a central server, and the method comprises: Determine the target sample type node to which the current sample belongs from the constructed detection domain hierarchical structure; the detection domain hierarchical structure includes at least one sample type node, at least one dimension node connected under each sample type node, and at least one detection method node connected under each dimension node; any node in the detection domain hierarchical structure is assigned a contribution weight for malicious sample analysis; For each dimension node connected to the target sample type node, based on the dimension feature required by the detection method indicated by each detection method node under the dimension node, the dimension feature data corresponding to the dimension feature is extracted from the current sample and sent to the detection engine running the detection method for detection to obtain the detection result corresponding to the current sample; the detection engine is assigned a participation weight for participating in malicious sample analysis of samples belonging to the target sample type; the target sample type refers to the sample type indicated by the target sample type node; Receive the detection result corresponding to the current sample reported by the detection engine; if the participation weight assigned to each detection engine that has reported the detection result meets the requirements of the set secret sharing method, then for each detection result reported by the detection engine, determine the contribution weight of the detection result based on the contribution weight assigned to the detection method node corresponding to the detection method run by the detection engine, the contribution weight assigned to the dimension node connected to the detection method node, and the contribution weight assigned to the target sample type node; aggregate each detection result according to the contribution weight of the detection result to analyze whether the current sample is a malicious sample based on the aggregated result.

2. The method according to claim 1, characterized in that The step of determining the contribution weight of each detection result reported by each detection engine based on the contribution weight assigned to the detection method node corresponding to the detection method run by the detection engine, the contribution weight assigned to the dimension node connected to the detection method node, and the contribution weight assigned to the target sample type node includes: For each detection result reported by the detection engine, the contribution weight assigned to the detection method node corresponding to the detection method run by the detection engine, the contribution weight assigned to the dimension node connected to the detection method node, and the contribution weight assigned to the target sample type node are performed by performing specified operations to obtain the operation result, and the operation result is determined as the contribution weight of the detection result.

3. The method according to claim 1, characterized in that: The sum of the participation weights of the detection engines running on the central server and participating in malicious sample analysis of samples belonging to the target sample type is less than the set minimum participation weight; The participation weights assigned to the detection engines that have reported the detection results meet the set secret sharing method requirements, which means that the sum of the participation weights assigned to the detection engines that have reported the detection results is greater than or equal to the set minimum participation weight.

4. The method according to claim 1, characterized in that: The method further comprises: For each detection engine, determine the loyalty of the detection engine based on the communication behavior between the detection engine and the central server, and / or determine the loyalty of the detection engine based on the detection results reported by the detection engine, and / or determine the loyalty of the detection engine based on the test results of the detection engine; Based on the loyalty of the detection engine, the participation weight of the detection engine is adjusted; wherein, the smaller the loyalty of the detection engine is, the smaller the participation weight of the detection engine is.

5. The method according to claim 4, characterized in that Determining the loyalty of the detection engine based on the communication behavior between the detection engine and the central server includes: When the detection engine is not located on the central server, the following judgment is performed: Determine whether the detection result reported by the detection engine is not received within a set time period after the dimensional feature data is sent to the detection engine; and / or, on the premise that the detection result reported by the detection engine is received within the set time period, determine whether the session between the detection engine and the central server established based on the participation weight of the detection engine is erroneous; wherein, when the central server receives the detection result reported by the detection engine through the encrypted channel established between the detection engine and the detection engine, if the decryption of the detection result based on the participation weight of the detection engine fails, it is determined that the session between the detection engine and the central server established based on the participation weight of the detection engine is erroneous; if the decryption is successful, it is determined that the session is successfully established; the encrypted channel is a secure channel constructed based on the participation weight of the detection engine as the channel key; The loyalty of the detection engine is determined based on the judgment result, wherein when the judgment result is yes, the loyalty of the detection engine is lower than when the judgment result is no.

6. The method according to claim 4, characterized in that Determining the loyalty of the detection engine based on the detection result reported by the detection engine includes: For the dimension node connected to the detection method node corresponding to the detection method run by the detection engine, based on the contribution weights assigned to the detection method nodes connected to the dimension node, the detection results reported by the detection engines that respectively run the detection methods corresponding to the detection method nodes are aggregated to obtain the aggregated sub-result corresponding to the dimension node; the loyalty of the detection engines that run the detection methods corresponding to the detection method nodes connected to the dimension node is determined based on the aggregated sub-result; wherein, the greater the difference between the detection result reported by any detection engine and the aggregated sub-result, the lower the loyalty of the detection engine; or, For each detection engine, the loyalty of the detection engine is determined based on the comparison result between the detection result reported by the detection engine and the converged result; wherein, the greater the difference between the detection result reported by any detection engine and the converged result, the lower the loyalty of the detection engine.

7. The method according to claim 4, characterized in that Determining the loyalty of the detection engine based on the test result of the detection engine includes: The detection engine is tested using pre-calibrated test samples, and the loyalty of the detection engine is determined based on the test results; wherein, the greater the difference between the test result of any test sample and the calibrated label of the test sample, the lower the loyalty of the detection engine.

8. The method according to any one of claims 4 to 7, characterized in that: The method further comprises: For each detection engine, if the participation weight of the detection engine is less than the set weight threshold, the detection engine is deregistered, and the detection engine is prohibited from registering with the central server again within a set time period after deregistration.

9. The method according to claim 1, characterized in that: The method further comprises: The analysis result of whether the current sample is a malicious sample is used as the sample label of the current sample and added to the original data set together with the current sample as labeled data; the detection method for optimizing and training each detection engine based on the original data set; and The contribution weight assigned to at least one node in the detection domain hierarchy is adjusted.

10. A multi-dimensional feature collaborative malicious sample analysis system, characterized in that: The system includes a central server, and at least one edge device independent of the central server; The central server executes the steps in any one of the methods described in claims 1 to 9; Any edge device loads at least one detection engine registered with the central server; any detection engine runs a detection method indicated by a detection method node, and the detection engine is assigned a participation weight for participating in malicious sample analysis of samples belonging to the target sample type; The target sample type refers to the sample type indicated by the target sample type node; the target sample type node is the sample type node to which the current sample belongs, determined by the central server from the constructed detection domain hierarchical structure according to the steps in any one of the methods described in claims 1 to 9.

11. A malicious sample analysis device with multi-dimensional feature collaboration, characterized in that: The device is applied to a central server, and the device comprises: A determination unit, used to determine the target sample type node to which the current sample belongs from a constructed detection domain hierarchical structure; the detection domain hierarchical structure includes at least one sample type node, at least one dimension node connected under each sample type node, and at least one detection method node connected under each dimension node; any node in the detection domain hierarchical structure is assigned a contribution weight for malicious sample analysis; and, For each dimension node connected to the target sample type node, based on the dimension feature required by the detection method indicated by each detection method node under the dimension node, the dimension feature data corresponding to the dimension feature is extracted from the current sample and sent to the detection engine running the detection method for detection processing to obtain the detection result corresponding to the current sample; the detection engine is assigned a participation weight for participating in malicious sample analysis of samples belonging to the target sample type; the target sample type refers to the sample type indicated by the target sample type node; An analysis unit is used to receive a detection result corresponding to a current sample reported by a detection engine; if the participation weight assigned to each detection engine that has reported the detection result meets the requirements of a set secret sharing method, then for each detection result reported by the detection engine, based on the contribution weight assigned to the detection method node corresponding to the detection method run by the detection engine, the contribution weight assigned to the dimension node connected to the detection method node, and the contribution weight assigned to the target sample type node, determine the contribution weight of the detection result; aggregate each detection result according to the contribution weight of the detection result to analyze whether the current sample is a malicious sample based on the aggregated result.